[{"_1":2,"_525":-5,"_526":-5},"loaderData",{"_3":4,"_30":31,"_522":523,"_524":-5},"root",{"_5":-5,"_6":7,"_10":11,"_18":17,"_19":-5,"_20":21,"_25":26},"theme","user",{"_8":9},"kind","anonymous","preferences",{"_12":13,"_14":15,"_16":17},"DOCS_SIDEBAR","DOCS_SIDEBAR_EXPANDED","PREDICTIONS_SIDEBAR","EXPANDED","DOCS_LANGUAGE","","webCsrfToken","principalToken","flags",{"_22":23,"_24":23},"showPipelinesUI",false,"showPylonWidget","env",{"_27":-7,"_28":29},"REPLICATE_API_BASE_URL","PYLON_APP_ID","0e09a6a8-a118-4139-a151-8813f39a36af","routes/changelog/layout",{"_32":33},"posts",[34,45,51,57,63,69,75,81,87,93,99,105,111,117,123,129,135,141,146,151,157,162,168,174,180,186,192,198,204,210,216,222,227,233,239,245,251,257,263,269,275,281,287,293,299,305,311,317,323,329,334,339,345,351,357,363,369,375,381,386,392,398,404,410,416,422,428,434,440,446,452,458,464,470,476,482,488,492,498,504,510,516],{"_35":36,"_37":38,"_39":40,"_41":42,"_43":44},"slug","2025-03-28-the-little-things-week-ending-march-28-2025","title","The little things, week ending March 28, 2025","publishedAt","2025-03-28T09:00:00Z","content","

API

models.examples.list - List example predictions made using the model
models.readme.get - Get the README content for a model

Cog

Cog 0.14 is out! The main new feature in 0.14.0 is models that run predictions concurrently via async/await. At time of writing the latest version is now 0.14.3.

Platform

Added directions for accessing docker images on private models

Playground

Fixed an issue that sometimes caused the playground to reset to the default inputs between runs
Fixed a bug when displaying multiple video inputs in the playground (beta)

Web

Overhauled navigation to focus on specific tasks
Improved feature flag implementation
Updated the fallback cover image for models
Fixed a bug that may have caused visual artifacts when using WebGL
Fixed a bug with z-indexes between navigation and search

","publishedAtDisplay","March 28, 2025",{"_35":46,"_37":47,"_39":48,"_41":49,"_43":50},"2025-03-14-the-little-things-week-ending-march-14-2025","The little things, week ending March 14, 2025","2025-03-14T09:00:00Z","

Web

Improvements to the org creation experience
Added better support for feature flags across the site
Improved the error messages for failed trainings
Fixed a bug with the sitemap for docs
Added a visual to make it clearer when a prediction has multiple outputs

Playground

Updated defaults to include new models
Fixed an issue when using multiple video files as inputs for a prediction

Docs

Build a webhook notifier with Val Town
Working with LoRAs

","March 14, 2025",{"_35":52,"_37":53,"_39":54,"_41":55,"_43":56},"2025-02-28-the-little-things-week-ending-february-28-2025","The little things, week ending February 28, 2025","2025-02-28T09:00:00Z","

Web

Launched better predictions list filtering
Launched the ability to create organizations without GitHub
Fixed a layout bug on the deployment overview
Tidied up the display of predictions on the deployment overview
Tweaked the display of changelog items
Fixed a bug when filtering dates in certain timezones and locales
Fixed a bug that caused certain output file types to fail to load on the prediction web view
Overhauled the account switcher to be more reliable
Updated the account view to separate out account and organization management
Overhauled the account update form view

Playground

Added the state of a playground prediction to the grid view

","February 28, 2025",{"_35":58,"_37":59,"_39":60,"_41":61,"_43":62},"2025-02-20-better-prediction-list","Better prediction list","2025-02-20T09:00:00Z","

$\"A$

You can now filter the last 24 hours of predictions by model, deployment, version, and status. You can also quickly scan through all your predictions by date.

Take a look at replicate.com/predictions.

","February 20, 2025",{"_35":64,"_37":65,"_39":66,"_41":67,"_43":68},"2025-02-14-the-little-things-week-ending-february-14-2025","The little things, week ending February 14, 2025","2025-02-14T09:00:00Z","

Web

Added list of recent predictions to the deployments overview
Fixed a bug when using the browser back button
Fix for trigger word highlighting when using certain characters in the playground
Minor updates to terms of service
Created the official models collection
Made the relationship between featured models and free usage clearer throughout the site

Playground

Added the ability to restore model defaults
Added a way to run multiple generations at once
Updated the UI to make it clearer when you can select multiple items
Fixed an error caused by losing connectivity while on the playground
Fixed a bug that made numeric inputs unclearable
Made the selected model persist on refresh
Improved the date picker visibility in light mode

Docs

Published a guide to building CI/CD pipelines for Cog models
Released a video showing how to run Replicate models on Cloudflare Workers
Updated the Getting started with Next.js guide and repo to use Flux Schnell
Update API docs with examples for fetching model input and output schemas.
Fixed a bug which could cause the sidebar to display on smaller breakpoints

API

Added created_before and created_after filters to the "list predictions" HTTP API

","February 14, 2025",{"_35":70,"_37":71,"_39":72,"_41":73,"_43":74},"2025-01-31-the-little-things-week-ending-january-31-2025","The little things, week ending January 31, 2025","2025-01-31T09:00:00Z","

Web

Launched official models
Added latest predictions to the deployment detail page
Fixed an issue with displaying certain pages on Replicate when logged out
Updated the model detail API playground to make it possible to train with inputs that aren’t .zip
Made it clearer when creating a deployment that flux fine-tunes shouldn’t be deployed
Fixed a bug where numeric inputs could increment on scroll on the model detail API playground
Fixed a bug that could cause inaccessible models to show up in collections
Improved the experience of sharing predictions via web
Improved clarity of autoscaling charts for deployments
Made it easier to copy the model identifier (for official models) or the model identifier and sha (for community models)
Updated chart for GPU metrics to make values easier to differentiate

Playground

Added an icon to thumbnails to differentiate video content from other types
Made it easier to see when errors occur and why
Added the ability to remove multiple items from the grid on the playground
Fixed a bug when selecting individual items that could cause all items to be deselected
Added Vercel AI SDK snippets to playground
Improved the loading and handling of large number of videos

Docs

Updated deployment documentation
Improved syntax highlighting across docs

","January 31, 2025",{"_35":76,"_37":77,"_39":78,"_41":79,"_43":80},"2025-01-29-official-models","Official models","2025-01-29T09:00:00.000Z","

Official models are always on and have predictable pricing. We maintain them in collaboration with the authors of the model to make sure they're high quality.

A number of models have worked this way for a while, but we're now giving it a name to make clear which ones work like this.

API

The way you call these models is a little different. If you're using a client library, you don't need to specify a version. For black-forest-labs/flux-1.1-pro in Node.js, for example:

const output = await replicate.run(\n\t"black-forest-labs/flux-1.1-pro",\n\t{ prompt: "A t-rex on a skateboard looking cool" }\n);

If you're using the HTTP API, you use the POST /models/<owner>/<name>/predictions endpoint and you don't need to specify a version. For example:

curl https://api.replicate.com/v1/models/black-forest-labs/flux-1.1-pro/predictions \\\n\t--request POST \\\n\t--header "Authorization: Bearer $REPLICATE_API_TOKEN" \\\n\t--header "Content-Type: application/json" \\\n\t--header "Prefer: wait" \\\n\t--data @- <<'EOM'\n{\n\t"input": {\n      "prompt": "A t-rex on a skateboard looking cool"\n\t}\n}\nEOM

Nothing has changed about how you run other models. The best way to find out how to run a model is the API documentation on a model.

Pricing

Instead of being charged by the amount of time a model runs, you're charged by output. For example, black-forest-labs/flux-1.1-pro is charged for each image it generates, but for other models this might be things like the number of tokens or the length of a video.

You can find out about each model's pricing on the pricing section of the model.

Models

Here are some of the models that are now official:

Take a look at the official models collection for the full list.

","January 29, 2025",{"_35":82,"_37":83,"_39":84,"_41":85,"_43":86},"2025-01-17-the-little-things-week-ending-january-17-2025","The little things, week ending January 17, 2025","2025-01-17T09:00:00Z","

Web

Relaunched the Replicate homepage
Fixed a bug that caused the Replicate site to display a flash of no styles during deployments
Improved the mobile navigation
Fixed rounding in deployment GPU memory chart
Migrated Remix to React Router 7
Fixed a bug that shows run costs as $null
Added video output previews to prediction table
Fixed an issue where videos took over the screen for mobile devices
Updated terms of service
Made it easier to open support tickets when you’re an organization without an email address set
Added a link to unpaid invoices in the dashboard to make it easier to pay them
Updated training form to allow more file types than just .zip

Playground

Added video models to the playground by default
Tweaked the UI elements on the playground to make interactivity clearer
Added deep linking for models, allowing them to be automatically added to the playground
Improved the visibility of impacted form fields when tweaking predictions on the playground
Updated downloaded files to use standardized filenames and zip files
Added keyboard shortcuts to prediction detail dialog to make it easier to flick through predictions
Updated messaging to make it clearer that any model with URI output is supported
Reordered models to show user-added ones first

Docs

Added markdown routes to all docs
Resolved an issue that made it difficult to navigate to site policy
Added a clearer way to get to the next part of an article

","January 17, 2025",{"_35":88,"_37":89,"_39":90,"_41":91,"_43":92},"2024-12-20-the-little-things-week-ending-december-20-2024","The little things, week ending December 20, 2024","2024-12-20T09:00:00Z","

Web

Launched GPU usage monitoring for deployments
Updated time display on the dashboard to be more granular
Fixed an issue where audio files weren't rendering correctly when provided as blobs
Added dark mode support for additional media types
Overhauled the model examples interface for model owners
Added the ability to test different lighting styles on the 3D renderer, for models like firtoz/trellis
Made it easier to view predictions by model version for model owners
Fixed a bug when linking to billing from disabled deployments
Dark mode fixes for table views
Made it clearer when sharing a prediction from a private model that it will make the prediction public
Better support loading deployment metric graphs with lots of data
Added the ability to click from deployment prediction graph to the relevant predictions
Fixed a bug that would cause the browser to record history push changes when toggling dark mode

Platform

Added support for GPU usage monitoring for deployments

Playground

Updated model list to be easier to browse for larger numbers of models

Docs

Published a guide to dynamic image caching with Cloudflare
Added a list of our subprocessors

","December 20, 2024",{"_35":94,"_37":95,"_39":96,"_41":97,"_43":98},"2024-12-17-language-model-training-no-longer-supported","Language model training no longer supported","2024-12-17T17:00:00Z","

We've stopped our experimental support for language model training.

You can still run all of the existing language models and fine-tuned models, but we no longer support training on those models to create derivative fine-tunes.

You can also still publish your own custom models using Cog.

Check out these docs for more information about building and publishing custom models:

We do still support fine-tuning image models. Check out these docs for more information:

","December 17, 2024",{"_35":100,"_37":101,"_39":102,"_41":103,"_43":104},"2024-12-13-gpu-memory-monitoring","GPU memory monitoring","2024-12-13T09:00:00Z","

$\"A$

You can now monitor how much GPU memory your deployments are using. The visualization shows the total memory available to your deployment, and a median and max usage over the last 2 or 24 hours across all your instances.

Take a look at replicate.com/deployments.

","December 13, 2024",{"_35":106,"_37":107,"_39":108,"_41":109,"_43":110},"2024-12-06-the-little-things-week-ending-december-06-2024","The little things, week ending December 06, 2024","2024-12-06T09:00:00Z","

Playground

Fixed some significant rendering bugs in Safari, especially on the model picker and grid view
Simplified the beta banner to be less obtrusive

Web

You can now view errors in your prediction or other lists inline
Fixed a bug that caused the browser navigation API to add multiple instances of each page
Collections contain featured models, hand picked by our team
The text-to-video collection now includes six of the top models from the Artificial Analysis leaderboard.
WebGL errors are better handled on the site, like when you see the Replicate goo
You can now deep link to default models on the playground (beta)
Fixed a bug when highlighting the trigger_word on Flux fine-tunes
Increased the number of featured models visible on the explore page
Improved sharing of collections and blog posts on social media
Improved the availability of the account switcher across Replicate
Added the status badge more consistently across Replicate
Fixed an issue with the blog RSS feed that caused images to not be displayed
Made it clearer that customers only pay for run time on private trained models
Updated model detail page to more accurately reflect the hardware being used when created through training

Docs

The text-to-video collection includes six of the top models from the Artificial Analysis leaderboard.

","December 6, 2024",{"_35":112,"_37":113,"_39":114,"_41":115,"_43":116},"2024-11-22-the-little-things-week-ending-november-22-2024","The little things, week ending November 22, 2024","2024-11-22T17:55:39Z","

Playground

Added FLUX1.1 [pro] ultra to default models, along with a number of new models
Improved image loading on playground
Optimizations when rendering many images, especially with larger images files (like with black-forest-labs/flux-1.1-pro-ultra)
Better handling of enums of numbers in the playground
Fixed a bug that cropped images in the grid
Fixed a bug where an image may have been prematurely flagged as downloadable
Added the ability to invert masks after inpainting, undo strokes, and to use the mouse wheel to alter brush resizing

Web

Improved signposting when you first sign up and land on the dashboard
Moved thumbnail images to the right in the prediction list to make browsing easier
Improved navigation rendering across Replicate
Added the ability to show preview pricing for hardware on pricing
Improved the onboarding flow to make the steps clearer
Improved invoice JSON download visibility
Improved discoverability of the new playground from model detail pages
Updated terms and privacy pages

Platform

Added support for per-video pricing
Added support for L40S GPUs

Docs

Published a new doc about OpenAPI schemas for the HTTP API and all models.

","November 22, 2024",{"_35":118,"_37":119,"_39":120,"_41":121,"_43":122},"2024-11-08-the-little-things-week-ending-november-8-2024","The little things, week ending November 8, 2024","2024-11-08T23:22:36Z","

Runtime

Rollback of data URLs in the new sync API.

Playground

Added file picker to playground to simplify file uploading
Added Stable Diffusion 3.5L to the default models
Added Recraft V3 to the default models
Made keyboard shortcuts clearer between different operating systems
Added download link to the grid in the playground
Fixed an issue where predictions were incorrectly being flagged as ephemeral
Added model card to the playground, to give more context on models

Web

Made keyboard shortcuts clearer between different operating systems
Improved the rendering and support of legacy blog posts
Updated API examples to better represent different return types
Made switching between themes easier across the site
Made the placement of “add to example” buttons for model owners on predictions clearer
Ongoing improvements to dark mode across web
Updated sitemaps for web
Resolved an issue where certain pages would redirect to add a trailing slash to the URL

Docs

Added HTTP API docs for new synchronous predictions.
Updated sitemaps for docs

","November 8, 2024",{"_35":124,"_37":125,"_39":126,"_41":127,"_43":128},"2024-11-07-a-fond-farewell-to-python-37","A fond farewell to Python 3.7","2024-11-07T16:07:24Z","

Replicate will stop supporting models built with Python 3.7 at 12:00 PM UTC (Coordinated Universal Time) on Friday, November 15th.

Python 3.7 was declared end-of-life in late June 2023. Since then, we've continued running models built with it to give model maintainers time to migrate to newer Python versions.

Although we would like to continue running these models indefinitely, we must prioritize the security needs of our customers. After Python 3.7 support ends, requests to affected models will fail with a 400-level error with an explanation.

Please update your models to a newer Python version before this date to avoid any interruptions, and let us know if you have questions or need help updating your models.

","November 7, 2024",{"_35":130,"_37":131,"_39":132,"_41":133,"_43":134},"2024-10-25-the-little-things-week-ending-2024-10-25","The little things, week ending October 25, 2024","2024-10-25T17:08:26Z","

Runtime

Dropped support for running Python 3.6 and 3.7 models
Dropped support for pushing Python 3.8 models, but existing ones will still run
Added support for Pydantic 2

Playground

Added inline inpainting to the playground
Added video to the playground
Added more granular grid sizes to the playground
Fixed a bug that caused double submissions on the playground form
Added better handling for unsupported outputs on the playground
Improved focus states in the grid and list view for the playground
Added the ability to delete items from the playground grid
Made keyboard shortcuts more visible for the playground
Improved scrolling performance for playground

Web

Relaunched the about page, including job roles and applications
Better handle longer queue lengths for deployment metrics
Fixed a display bug with pricing on models for signed out users
Fixed some z-index weirdness for signed out users
Improved copy / saving of streamed images on web
Fixed bugs in per-model API endpoint documentation
Handled models without a cover image better, especially on social media
Reduced inconsistencies for input schema tables between model versions and the model API tab
Improved display of plural units in the model pricing table
Added sorting by name and cost to invoices, and the ability to download invoices as JSON
Improved handling of multi file strings on the model page
Fixed bug on the model detail page when pasting JSON
Handled an edge case where the state of a model didn’t match visually
Fixed a few cases where the colors were wrong in dark mode

Docs

Moved guides into docs
Updated selection color on our docs to improve accessibility
Improved readability for monospaced fonts on docs
Handled image caching for docs better

","October 25, 2024",{"_35":136,"_37":137,"_39":138,"_41":139,"_43":140},"2024-10-10-dark-mode","Dark mode","2024-10-10T00:50:00Z","

$\"A$

Finally. You can switch it in the footer.

","October 10, 2024",{"_35":142,"_37":143,"_39":144,"_41":145,"_43":140},"2024-10-10-playground-beta","Playground (beta)","2024-10-10T00:49:52Z","

Playground is a way to quickly try out and compare the output of models on Replicate.

$\"A$

You can:

Create grids of images and compare them
Grab the code to generate the images with Replicate
Test different models, prompts, and settings, side-by-side
Keep a scrapbook and export images in bulk

Playground is currently in beta, and works with any FLUX and related fine-tunes. Try it out at replicate.com/playground.

We'd love to hear what you think. Send your feedback to playground@replicate.com.

",{"_35":147,"_37":148,"_39":149,"_41":150,"_43":140},"2024-10-10-new-documentation","New documentation","2024-10-10T00:13:01Z","

$\"A$

Documentation now has a cleaner design, better navigation, and lots of new content to help you build with Replicate.

It’s organized into four sections:

Getting started: tutorials to help you get up and running quickly
Guides: step-by-step instructions and deeper dives into specific topics
Topics: detailed explanations of how Replicate works
Reference: complete documentation for Replicate’s client libraries and API

We've added dozens of new code samples, detailed guides for common use cases, and an expanded API reference.

Take a look at replicate.com/docs.

",{"_35":152,"_37":153,"_39":154,"_41":155,"_43":156},"2024-10-09-synchronous-api","Synchronous API","2024-10-09T22:56:59Z","

Our client libraries and API are now much faster at running models, particularly if a file is being returned.

The API now just returns the response immediately. Before, you would have to poll to get the result.

If you're using the Node.js or Python client libraries, you don't have to worry about this. Just upgrade to the latest version and it gets faster. Also, instead of returning HTTP URLs, they now return file objects, which makes it much easier to write them to storage or pass to HTTP responses.

Node.js

Install the beta version of the client library:

npm install replicate@latest

Then run the model:

import Replicate from "replicate";\nimport fs from "node:fs";\n \nconst replicate = new Replicate();\nconst [output] = await replicate.run("black-forest-labs/flux-schnell", { input: { \n\tprompt: "astronaut riding a rocket like a horse"\n}});\n \n// It now returns a file object\nfs.writeFileSync("my-image.webp", output);\n \n// Or, you can still get an HTTP URL\nconsole.log(output.url())

Python

Update the client library:

pip install --upgrade replicate

Then run the model:

[output] = replicate.run(\n  "black-forest-labs/flux-schnell",\n  input={"prompt": "astronaut riding a rocket like a horse"}\n);\n \nwith open('output.webp', 'wb') as file:\n    file.write(output.read())\n \nprint(output.url)

The client libraries returning file objects is a breaking change so be careful when you upgrade your apps.

HTTP API

If you're using the HTTP API for models or deployments, you can now pass the header Prefer: wait, which will keep the connection open until the prediction has finished:

curl -X POST <https://api.replicate.com/v1/models/black-forest-labs/flux-schnell/predictions> \\\\\n     -H "Authorization: Bearer $REPLICATE_API_TOKEN" \\\\\n     -H "Prefer: wait=10" \\\\\n     -d '{"input": {"prompt": "a cat riding a narwhal with rainbows"}}'

Output:

{\n  "id": "dapztkbwgxrg20cfgsmrz2gm38",\n  "status": "processing",\n  "output": ["https://..."],\n}

By default it will wait 60 seconds before returning the in-progress prediction. You can adjust that by passing a time, like Prefer: wait=10 to wait 10 seconds.

Take a look at the docs on creating a prediction for more details.

","October 9, 2024",{"_35":158,"_37":159,"_39":160,"_41":161,"_43":156},"2024-10-09-large-log-outputs-might-be-truncated","Large log outputs might be truncated","2024-10-09T10:24:14Z","

Predictions or trainings that log particularly large volumes of information may now have their logs truncated. This helps us manage load on our platform and is a restriction we may lift in future.

For now, you'll always get the first few lines of your logs -- which might contain important information about inputs, seed values, etc. -- and the most recent logs.

",{"_35":163,"_37":164,"_39":165,"_41":166,"_43":167},"2024-09-17-time-limit-for-sharing-predictions","Time limit for sharing predictions","2024-09-17T19:06:33Z","

To improve API performance and speed up database queries, we are cleaning out old\npredictions.

From now on, you'll only be able to share predictions made in the web playground for up to\n60 days after they're created.

If you've already shared a prediction, don't worry - it will still be accessible to others\neven after 60 days.

","September 17, 2024",{"_35":169,"_37":170,"_39":171,"_41":172,"_43":173},"2024-09-11-view-training-outputs-on-the-web","View training outputs on the web","2024-09-11T22:35:28Z","

The training detail page on the website now has a JSON tab, so you can see all the metadata and outputs from your trainings, including the URL of your generated weights file.

To view all your trainings, go to replicate.com/trainings

$\"JSON$

","September 11, 2024",{"_35":175,"_37":176,"_39":177,"_41":178,"_43":179},"2024-07-24-api-for-searching-public-models","API for searching public models","2024-07-24T15:15:41Z","

Replicate’s API now has an endpoint for searching public models.

To use it, make a QUERY HTTP request using your search query as the plaintext body of the request:

curl -s -X QUERY \\\n  -H "Authorization: Bearer $REPLICATE_API_TOKEN" \\\n  -H "Content-Type: text/plain" \\\n  -d "face" \\\n  https://api.replicate.com/v1/models

The response will be a paginated JSON object containing an array of model objects:

{\n  "next": null,\n  "previous": null,\n  "results": [\n    {\n      "url": "https://replicate.com/acme/hello-world",\n      "owner": "acme",\n      "name": "hello-world",\n      "description": "A tiny model that says hello",\n      "visibility": "public",\n      "github_url": "https://github.com/replicate/cog-examples",\n      "paper_url": null,\n      "license_url": null,\n      "run_count": 5681081,\n      "cover_image_url": "...",\n      "default_example": {...},\n      "latest_version": {...}\n    }\n  ]\n}

For more details, check out the HTTP API reference docs.

You can also the download metadata for all public models on Replicate using the list public models API if you need more fine-grained control or want to build your own implementation of model search.

","July 24, 2024",{"_35":181,"_37":182,"_39":183,"_41":184,"_43":185},"2024-07-15-streams-always-available-stream-parameter-deprecated","Streams always available, stream parameter deprecated","2024-07-15T16:04:02Z","

We've made streaming a tiny bit easier and simplified our API.

When we first launched streaming for language models you had to pass "stream": true with your request to enable it. Adoption of streaming has been great, with about one third of requests to compatible models requesting a stream, and our streaming infrastructure is more robust now than it was when we first launched it. This means we're now in the position to just stream everything by default if the model supports it.

So you no longer need to pass "stream": true with your prediction request to get back a stream URL – if the model supports streaming, you'll always get one back.

We've deprecated the field, but it's still valid to pass it. If you do continue to pass "stream": true then if the model supports streaming it has no effect. And if the model doesn't support streaming we'll keep the current behaviour of responding with a 422.

","July 15, 2024",{"_35":187,"_37":188,"_39":189,"_41":190,"_43":191},"2024-06-07-secret-inputs-for-models","Secret inputs for models","2024-06-07T17:17:17Z","

We've added a way to securely pass sensitive values to models.

In Cog v0.9.7 and later,\r\nyou can annotate an input with the Secret type\r\nto signify that an input holds sensitive information,\r\nlike a password or API token.

from cog import BasePredictor, Secret\n \n \nclass Predictor(BasePredictor):\n    def predict(self, api_token: Secret) -> None:\n        # Prints '**********'\n        print(api_token)        \n \n        # Use get_secret_value method to see the secret's content.\n        print(api_token.get_secret_value())

Replicate treats secret inputs differently throughout its system.\r\nWhen you create a prediction on Replicate,\r\nany value passed to a Secret input is redacted after being sent to the model.

Before

$\""A$

After

$\""This$

Caution: Passing secret values to untrusted models can result in unintended disclosure, exfiltration, or misuse of sensitive data.

","June 7, 2024",{"_35":193,"_37":194,"_39":195,"_41":196,"_43":197},"2024-06-04-disable-api-tokens","Disable API tokens","2024-06-04T15:48:06Z","

You can now disable API tokens from the web. This is useful if you accidentally leaked your token and want to prevent unauthorized use of the token.

To view and manage your tokens, go to replicate.com/account/api-tokens.

$\"\"/$

","June 4, 2024",{"_35":199,"_37":200,"_39":201,"_41":202,"_43":203},"2024-05-28-rss-and-atom-feeds","RSS and Atom feeds","2024-05-28T21:05:45Z","

We now publish RSS and Atom feeds for our blog, changelog, and platform status site. You can subscribe to these feeds to get updates about new product features, as well as platform incidents

If you're using Slack, you can use Slack's RSS app to subscribe to these feeds and get notifications right in your team's Slack workspace.

","May 28, 2024",{"_35":205,"_37":206,"_39":207,"_41":208,"_43":209},"2024-05-21-delete-stuff","Delete stuff","2024-05-21T23:45:49Z","

You can now delete models, versions, and deployments using the web or the HTTP API.

Deleting models

You can delete a model directly from the web on the model settings page. You can also delete a model programmatically using the models.delete HTTP API.

There are some restrictions on which models you can delete:

You can only delete models you own.
You can only delete private models.
You can only delete models that have no versions associated with them. You'll need to delete any versions first before deleting the model.

Deleting model versions

You can delete a model version directly from the web using the nav in the header of the model versions page. You can also delete a model version programmatically using the versions.delete HTTP API.

The following restrictions apply to deleting versions:

You can only delete versions from models you own.
You can only delete versions from private models.
You cannot delete a version if someone other than you has run predictions with it.
You cannot delete a version if it is being used as the base model for a fine tune/training.
You cannot delete a version if it has an associated deployment.
You cannot delete a version if another model version is overridden to use it.

Deleting deployments

You can delete a deployment directly from the web on the deployment settings page. You can also delete a deployment programmatically using the deployments.delete HTTP API.

The following restriction applies to deleting deployments:

You can only delete deployments that have been offline and unused for at least 15 minutes.

","May 21, 2024",{"_35":211,"_37":212,"_39":213,"_41":214,"_43":215},"2024-05-14-docs-and-client-library-support-for-webhook-verification","Docs and client library support for webhook verification","2024-05-14T23:36:35Z","

A webhook is an HTTP POST from an unknown source. Attackers can impersonate services by simply sending a fake webhook to an endpoint. Replicate protects you from this attack by signing every webhook and its metadata with a unique key for each user or organization. You can use this signature to verify that incoming webhooks are coming from Replicate before you process them.

Today we improved our documentation and client library support for webhooks, so you can securely verify webhooks in your existing application without having to write your own validation logic:

","May 14, 2024",{"_35":217,"_37":218,"_39":219,"_41":220,"_43":221},"2024-05-02-search-for-deployments","Search for deployments","2024-05-02T21:47:32Z","

The Replicate website has a search bar at the top of every screen that lets you fuzzy search for models, collections, and pages.

We've now added deployments to the search results, so you can quickly jump right to one of your deployments with fewer clicks.

Press cmd-k (or ctrl-k on Windows) to open the search and search for your deployment by name:

$\"\"/$

","May 2, 2024",{"_35":223,"_37":224,"_39":225,"_41":226,"_43":221},"2024-05-02-webhooks-activity-ui","Webhooks activity UI","2024-05-02T14:46:53Z","

We've added two new UI features to make your experience with webhooks better on Replicate.

On the prediction detail screen, if your prediction was made with webhooks, you'll see an indicator that opens a sliding panel with the details and status of those webhooks.

There's also a new "Webhooks" tab on the Dashboard where you'll find a chart detailing 6 hours of recent webhook activity, as well as any failures that have occurred in that time period.

",{"_35":228,"_37":229,"_39":230,"_41":231,"_43":232},"2024-05-01-improved-validation-for-api-prediction-payloads","Improved validation for API prediction payloads","2024-05-01T22:56:22Z","

If you typo a field in the JSON request body when creating a prediction with the API, we will now tell you about it rather than silently accepting it.

Here's an example showing the response you'll get if you misspell input as inptu:

$ curl -s -X POST \\\n  -H "Authorization: Bearer $REPLICATE_API_TOKEN" \\\n  -H 'Content-Type: application/json' \\\n  -d '{\n    "version": "5c7d5dc6dd8bf75c1acaa8565735e7986bc5b66206b55cca93cb72c9bf15ccaa", \n    "inptu": {"text": "Alice"}\n  }' \\\n  https://api.replicate.com/v1/predictions | jq .invalid_fields

[\n  {\n    "type": "required",\n    "field": "",\n    "description": "input is required"\n  },\n  {\n    "type": "additional_property_not_allowed",\n    "field": "",\n    "description": "Additional property inptu is not allowed"\n  }\n]

Note that this validation only applies to top-level properties in the payload like input, version, stream, webhook, webhook_events_filter, etc.

","May 1, 2024",{"_35":234,"_37":235,"_39":236,"_41":237,"_43":238},"2024-04-11-t4-models-now-have-more-ram","T4 models now have more RAM","2024-04-11T17:08:39Z","

Until recently, models running on T4 GPUs would get 4 CPUs, 8 GB of RAM and 16 GB of GPU RAM. We got feedback from model authors that this caused issues with loading more than 8GB of weights into the GPU - it's possible, but it requires writing more intricate code to stream weights from disk to GPU RAM without loading the whole weights into main RAM at once.

Starting today, T4 models can now use up to 16 GB of RAM. This should make it easier to make use of all of the GPU RAM on T4 GPUs. There is no pricing change for this upgrade.

You can see the resources available for each hardware type on our pricing page.

","April 11, 2024",{"_35":240,"_37":241,"_39":242,"_41":243,"_43":244},"2024-04-10-github-secret-scanning","GitHub secret scanning","2024-04-10T20:45:27Z","

When working with secrets like API tokens, you should always store them in your app configuration (like environment variables) rather than directly in your codebase. Keeping sensitive information separate from your source code prevents accidental exposure.

This is a common and well-known security practice, but sometimes you slip up and accidentally include a secret in your codebase. That's where GitHub security scanning comes in.

We've partnered with GitHub and are now members of their Secret scanning partner program, helping to keep your account safe. Whenever GitHub finds a Replicate API token in a public repository, they notify us so we can take proper steps to protect your account from unwanted use.

","April 10, 2024",{"_35":246,"_37":247,"_39":248,"_41":249,"_43":250},"2024-04-03-bearer-tokens","Bearer tokens","2024-04-03T21:14:58Z","

When using an HTTP API, a Bearer token is like a password that you specify in the Authorization header of your HTTP requests to access protected resources.

We’ve updated our API to support a standard Bearer authorization scheme rather than a custom/non-standard Token scheme. We've also updated our API docs and OpenAPI specification to use the new format:

curl -H "Authorization: Bearer r8_...." https://api.replicate.com/v1/models/replicate/hello-world

The old Authorization: Token r8... format is still supported, so your existing API integrations will continue to work without any code changes.

","April 3, 2024",{"_35":252,"_37":253,"_39":254,"_41":255,"_43":256},"2024-03-12-deployments-api","Deployments API","2024-03-12T04:59:29Z","

Deployments give you more control over how your models run. You can scale them up and down based on demand, customize their hardware, and monitor performance and predictions without editing your code.

Managing deployments was previously only possible on the web, but you can now also create, read, and update deployments using Replicate’s HTTP API.

Here's an example API request that updates the number of min and max instances for an existing deployment:

curl -s \\\n  -X PATCH \\\n  -H "Authorization: Token $REPLICATE_API_TOKEN" \\\n  -H "Content-Type: application/json" \\\n  -d '{"min_instances": 3, "max_instances": 10}' \\\n  https://api.replicate.com/v1/deployments/acme/my-app-image-generator

Check out the API docs for deployments:

If you're new to deployments, check out the getting started guide.

","March 12, 2024",{"_35":258,"_37":259,"_39":260,"_41":261,"_43":262},"2024-02-19-3d-viewer-for-glb-outputs","3D viewer for GLB outputs","2024-02-19T18:53:02Z","

Models like adirik/imagedream that output GLB files now have a web-based viewer, so you can explore the 3D output right in your browser.

The GLB format encapsulates textures, models, and animations into a single file which can be easily shared and used across platforms and devices. This makes it useful in web development, augmented reality projects, and game development.

If you're working with AI models that produce PLY outputs, you can use a model like camenduru/lgm-ply-to-glb to convert PLY files to GLB.

There's also a collection of models for making 3D stuff.

","February 19, 2024",{"_35":264,"_37":265,"_39":266,"_41":267,"_43":268},"2024-02-16-task-oriented-collections","Task-oriented collections","2024-02-16T20:04:37Z","

We've updated our curated model collections to be more oriented around tasks like upscaling images, generating embeddings, or getting structured data from language models. This should make it easier to find the right models for the problems you're trying to solve.

See replicate.com/collections

Each collection now also includes a more detailed summary of the kinds of tasks you can perform with models in that collection. For example, vision models can be used for all sorts of tasks like captioning images, answering questions about images, or detecting objects:

","February 16, 2024",{"_35":270,"_37":271,"_39":272,"_41":273,"_43":274},"2024-02-06-webhook-verification","Webhook verification","2024-02-06T23:44:07Z","

Webhooks provide real-time updates about your predictions. You can specify an endpoint when you create a prediction, and Replicate will send HTTP POST requests to that URL when the prediction is created, updated, and finished.

Starting today, we now sign every webhook and its metadata with a unique key for each user or organization. This allows you to confirm the authenticity of incoming webhooks and protect against unauthorized access and replay attacks.

Learn how to verify incoming webhooks on our webhooks guide

","February 6, 2024",{"_35":276,"_37":277,"_39":278,"_41":279,"_43":280},"2024-01-19-fine-tune-sdxl-from-the-web","Fine-tune SDXL from the web","2024-01-19T01:17:17Z","

In August 2023, we added an API for fine-tuning SDXL, Stability AI's most powerful open-source image generation model. We made it easy to create your own custom models with just a handful of training images and a few lines of code, then generate new images in your own style. In the last few months, people have created lots of interesting SDXL fine-tunes.

You can now create SDXL fine-tunes from the web. Choose a name for your model, drag and drop a zip file of images, and kick off the training process, without writing any code.

To get started, visit replicate.com/stability-ai/sdxl/train

$\"form\"/$

","January 19, 2024",{"_35":282,"_37":283,"_39":284,"_41":285,"_43":286},"2023-12-19-compare-image-inputs-to-outputs","Compare image inputs to outputs","2023-12-19T00:47:30Z","

We've added a new slider component that makes it easier to compare image inputs to outputs. Drag the slider to the left and right to compare the "before and after" images.

This feature works for all models that take an input image and produce an output image, like image restoration models and upscalers.

$\"slider\"/$

","December 19, 2023",{"_35":288,"_37":289,"_39":290,"_41":291,"_43":292},"2023-11-08-code-snippets-for-every-prediction","Code snippets for every prediction","2023-11-08T22:06:29Z","

Every prediction on the web now includes code snippets for Node.js, Python, and other languages. This makes it easy to tinker with a model on the web, then grab some useable code and run with it.

To view these code snippets, click the language tabs under the Input heading on any model or prediction page:

Clicking one of these language tabs changes the page URL, which means you can also deep link to language-specific snippets for any model:

Here's a five-minute video showing off how to use this new feature to run a model in the browser, then grab the code to run the model in your own project using Replicate's API:

\n","November 8, 2023",{"_35":294,"_37":295,"_39":296,"_41":297,"_43":298},"2023-11-06-api-for-creating-models","API for creating models","2023-11-06T19:34:02Z","

Replicate's API now has an endpoint for creating models.

You can use it to automate the creation of models, including fine-tunes of SDXL and Llama 2.

cURL usage

Here's an example that uses cURL to create a model with a given owner, name, visibility, and hardware:

curl -s -X POST -H "Authorization: Token $REPLICATE_API_TOKEN" \\\n    -d '{"owner": "my-username", "name": "my-new-model", "visibility": "public", "hardware": "gpu-a40-large"}' \\\n    https://api.replicate.com/v1/models

The response is a JSON object of the created model:

{\n  "url": "https://replicate.com/my-username/my-new-model",\n  "owner": "my-username",\n  "name": "my-new-model",\n  "description": null,\n  "visibility": "public",\n  "github_url": null,\n  "paper_url": null,\n  "license_url": null,\n  "run_count": 0,\n  "cover_image_url": null,\n  "default_example": null,\n  "latest_version": null\n}

To see all the hardware available for your model to run, consult our endpoint for listing hardware.

curl -s -H "Authorization: Token $REPLICATE_API_TOKEN" \\\n    https://api.replicate.com/v1/hardware

[\n  { "name": "CPU", "sku": "cpu" },\n  { "name": "Nvidia T4 GPU", "sku": "gpu-t4" },\n  { "name": "Nvidia A40 GPU", "sku": "gpu-a40-small" },\n  { "name": "Nvidia A40 (Large) GPU", "sku": "gpu-a40-large" }\n]

To compare the price and specifications of these hardware types, check out the pricing page.

JavaScript usage

We've added this new operation to the Replicate JavaScript client:

npm install replicate@latest

Then:

import Replicate from "replicate";\nconst replicate = new Replicate();\n \n// create a new model\nconst model = await replicate.models.create(\n    "my-username", \n    "my-new-model", \n    {\n        visibility: "public",\n        hardware: "gpu-a40-large"\n    }\n);\nconsole.log(model)

Python usage

We've added this new operation to the Replicate Python client:

pip install --upgrade replicate

Then:

import replicate\n \nmodel = replicate.models.create(\n    owner="my-username",\n    name="my-new-model",\n    visibility="public",\n    hardware="gpu-a40-large",\n)\nprint(model)

Elixir usage

We've added this new operation to the Replicate Elixir client:

mix deps.update replicate

Then:

iex> {:ok, model} =\n         \tReplicate.Models.create(\n        \t\towner: "your-username",\n        \t\tname: "my-model",\n         \t\tvisibility: "public",\n\t\t        hardware: "gpu-a40-large"\n       \t\t)

API docs

Check out the HTTP API reference for more detailed documentation about this new endpoint.

","November 6, 2023",{"_35":300,"_37":301,"_39":302,"_41":303,"_43":304},"2023-10-17-improved-training-detail-pages","Improved training detail pages","2023-10-17T18:09:56Z","

When you kick off a training process to fine-tune your own model, there's a page you can visit to view the status of the training, as well as the inputs and outputs. We've made some recent improvements to those pages:

The header now shows the base model and destination model you used when fine-tuning.
Metadata about the training job is now diplayed above the inputs and outputs to make it easier to see the important details without having to scroll down the page.
We've added code snippets you can copy and paste to run your new fine-tuned model.

$\"\"/$

","October 17, 2023",{"_35":306,"_37":307,"_39":308,"_41":309,"_43":310},"2023-10-09-prediction-parameters-as-json","Prediction parameters as JSON","2023-10-09T18:06:08Z","

You can now view prediction parameters as JSON from the prediction detail page.

This improves the workflow for experimenting in the web interface and then transitioning to making predictions from the API, using code.

$\"\"/$

","October 9, 2023",{"_35":312,"_37":313,"_39":314,"_41":315,"_43":316},"2023-10-05-api-for-listing-public-models","API for listing public models","2023-10-05T23:25:53Z","

Replicate's API now has an endpoint for listing public models.

You can use it to discover newly published models, and to build your own tools for exploring Replicate's ecosystem of thousands of open-source models.

cURL usage

Here's an example that uses cURL and jq to fetch the URLs of the 25 most recently updated public models:

 curl -s -H "Authorization: Token $REPLICATE_API_TOKEN" \\\n    https://api.replicate.com/v1/models | jq ".results[].url"

The response is a paginated JSON array of model objects:

{\n  "next": null,\n  "previous": null,\n  "results": [\n    {\n      "url": "https://replicate.com/replicate/hello-world",\n      "owner": "replicate",\n      "name": "hello-world",\n      "description": "A tiny model that says hello",\n      "visibility": "public",\n      "github_url": "https://github.com/replicate/cog-examples",\n      "paper_url": null,\n      "license_url": null,\n      "run_count": 5681081,\n      "cover_image_url": "...",\n      "default_example": {...},\n      "latest_version": {...}\n    }\n  ]\n}

JavaScript usage

We've added this new operation to the Replicate JavaScript client:

npm install replicate@latest

Then:

import Replicate from "replicate";\nconst replicate = new Replicate();\n \n// get recently published models\nconst latestModels = await replicate.models.list();\nconsole.log(latestModels)\n \n// paginate and get all models\nconst allModels = []\nfor await (const batch of replicate.paginate(replicate.models.list)) {\n  allModels.push(...batch);\n}\nconsole.log(allModels)

Python usage

We've added this new operation to the Replicate Python client:

pip install --upgrade replicate

Then:

import replicate\n \nmodels = replicate.models.list()\nprint(models)

Making your own models discoverable

If you’re deploying your own public models to Replicate and want others to be able to discover them, make sure they meet the following criteria:

The model is public.
The model has at least one published version.
The model has at least one example prediction. To add an example, create a prediction using the web interface then click the “Add to examples” button below the prediction output.

","October 5, 2023",{"_35":318,"_37":319,"_39":320,"_41":321,"_43":322},"2023-10-03-deployments","Deployments","2023-10-03T16:27:59Z","

You can now create a deployment to get more control over how your models run. Deployments allow you to run a model with a private, fixed API endpoint. You can configure the version of the model, the hardware it runs on, and how it scales.

Using deployments, you can:

Roll out new versions of your model without having to edit your code.
Keep instances always on to avoid cold boots.
Customize what hardware your models run on.
Monitor whether instances are booting up, running, or processing predictions.
View predictions that are flowing through your models.

Deployments work with both public models and your own private models.

$\"\"/$

🚀 Check out the deployments guide to learn more and get started.

","October 3, 2023",{"_35":324,"_37":325,"_39":326,"_41":327,"_43":328},"2023-10-02-prediction-query-parameter","Prediction query parameter","2023-10-02T20:14:35Z","

When you create a prediction on the web, we now append a query parameter ?prediction=<uuid>, so that if you refresh the page, you see that prediction instead of the form prefilled with the default inputs. Previously, if you created a prediction on the web and refreshed, you'd lose the prediction and have to go spelunking for it on your dashboard.

$\"query-param\"/$

","October 2, 2023",{"_35":330,"_37":331,"_39":332,"_41":333,"_43":328},"2023-10-02-fullscreen-training-logs","Fullscreen training logs","2023-10-02T20:05:29Z","

You can now expand your training logs and view them full-screen.

$\"click-fullscreen-logs\"/$

$\"fullscreen-logs\"/$

",{"_35":335,"_37":336,"_39":337,"_41":338,"_43":328},"2023-10-02-dynamic-status-favicons","Dynamic status favicons","2023-10-02T20:01:33Z","

✅ We've added a new feature to show the prediction status in the favicon of the browser tab. This makes it easier to know when your running predictions have completed without having to switch tabs.

$\"favicons\"/$

",{"_35":340,"_37":341,"_39":342,"_41":343,"_43":344},"2023-08-14-streaming-output-for-language-models","Streaming output for language models","2023-08-14T20:48:34Z","

Replicate's API now supports server-sent event (SSE) streams for language models, giving you live output as the model is running. See the announcement blog post and our streaming guide for more details about how to use streaming output.

$\"eventsource$

","August 14, 2023",{"_35":346,"_37":347,"_39":348,"_41":349,"_43":350},"2023-08-10-multiple-api-tokens-for-users","Multiple API tokens for users","2023-08-10T05:03:48Z","

You can now create multiple personal API tokens at https://replicate.com/account/api-tokens

They're just like organization tokens, but they're only for your personal user account. You can name your tokens to make them distinguishable, and reset them if needed.

$\"multiple-tokens\"/$

","August 10, 2023",{"_35":352,"_37":353,"_39":354,"_41":355,"_43":356},"2023-07-24-a40-gpus-now-available","A40 GPUs now available","2023-07-24T22:29:08Z","

You can now run Replicate models on NVIDIA A40 GPUs. In terms of price and performance, the A40 sits between our T4 and A100 hardware. For many models, the A40 Large can be 80-90% as fast as A100s but almost half the price.

The A40 GPU is currently available in two configurations, each with the same GPU but attached to a machine with different amounts of CPU and RAM:

Hardware	Price	GPU	CPU	GPU RAM	RAM
Nvidia A40 GPU	$0.0013 per second ($0.078 per minute)	1x	4x	48GB	16GB
Nvidia A40 (Large) GPU	$0.0016 per second ($0.096 per minute)	1x	10x	48GB	72GB

To choose which GPU type is used to run your model, see the Hardware dropdown on your model's settings page:

\n $\"select-gpu\"$ \n

To compare price and performance of all available GPUs, see our pricing page.

","July 24, 2023",{"_35":358,"_37":359,"_39":360,"_41":361,"_43":362},"2023-07-21-hardware-and-pricing-for-trainable-models","Hardware and pricing for trainable models","2023-07-21T22:38:39Z","

Trainable model pages now have a section that indicates training hardware type and cost.

See https://replicate.com/a16z-infra/llama7b-v2-chat#training

$\"training\"/$

","July 21, 2023",{"_35":364,"_37":365,"_39":366,"_41":367,"_43":368},"2023-07-20-training-api-for-language-models","Training API for language models","2023-07-20T01:41:29Z","

We built a training API for fine-tuning language models, and today we're making it available to all Replicate users.

You can fine-tune language models to make them better at a particular task, like classifying text, answering questions about your private data, being a chatbot, or extracting structured data from text.

You can train models like LLaMA 2, Flan T5, GPT-J and others. Check out the trainable language models collection to see what models can be fine-tuned, and stay tuned as we add support for more.

To get started, check out our guide to fine tuning a language model.

training = replicate.trainings.create(\n  version="a16z-infra/llama7b-v2-chat:a845a72bb3fa3ae298143d13efa8873a2987dbf3d49c293513cd8abf4b845a83",\n  input={\n    "train_data": "https://example.com/my-training-data.jsonl",\n  },\n  destination="zeke/my-custom-llama-2"\n)

","July 20, 2023",{"_35":370,"_37":371,"_39":372,"_41":373,"_43":374},"2023-05-19-git-commit-and-tag-for-model-versions","Git commit and tag for model versions","2023-05-19T19:10:19Z","

You can now see the Git commit and tag used to create new versions of a model. Run brew upgrade cog to upgrade to the latest release.

The next time you run cog push, the current Git commit and tag for your model will automatically be included in the resulting Docker image, and it'll show up on Replicate under the model page's Versions tab.

$\"git-commit\"/$

","May 19, 2023",{"_35":376,"_37":377,"_39":378,"_41":379,"_43":380},"2023-05-16-downloading-outputs","Downloading outputs","2023-05-16T19:00:48Z","

We've made a few improvements to the download mechanism on the website:

If you ran a model that produces multiple outputs, you can download them all as a zip file.
If you're downloading a single output, it will now actually download instead of linking to the file's URL.

Happy downloading!

$\"zip-downloads\"/$

","May 16, 2023",{"_35":382,"_37":383,"_39":384,"_41":385,"_43":380},"2023-05-16-invoice-breakdowns","Invoice breakdowns","2023-05-16T18:49:35Z","

You can now view a detailed summary of your invoices with a breakdown of cost, prediction time, and hardware for each model you ran.

\n $\"Example$ \n

Check out your billing settings at https://replicate.com/account/billing

",{"_35":387,"_37":388,"_39":389,"_41":390,"_43":391},"2023-05-09-swift-client-library","Swift client library","2023-05-09T20:05:10Z","

Replicate now has a client library for Swift. It’s got everything you need to build an AI-powered app for iOS and macOS using Replicate's HTTP API.

Add it as a package dependency to your project:

let package = Package(\n    // name, platforms, products, etc.\n    dependencies: [\n        // other dependencies\n        .package(url: "https://github.com/replicate/replicate-swift", from: "0.12.1"),\n    ],\n    targets: [\n        .target(name: "<target>", dependencies: [\n            // other dependencies\n            .product(name: "Replicate", package: "replicate-swift"),\n        ]),\n        // other targets\n    ]\n)

Then, you can run predictions:

import Replicate\n \n// Get your API token at https://replicate.com/account\nprivate let replicate = Replicate.Client(token: <#token#>)\n \nlet output = try await replicate.run(\n    "stability-ai/stable-diffusion:db21e45d3f7023abc2a46ee38a23973f6dce16bb082a930b0c49861f96d1e5bf",\n    ["prompt": "a 19th century portrait of a wombat gentleman"]\n)\n \nprint(output)\n// ["https://replicate.com/api/models/stability-ai/stable-diffusion/files/50fcac81-865d-499e-81ac-49de0cb79264/out-0.png"]

Follow our guide to building a SwiftUI app or read the full documentation on GitHub.

","May 9, 2023",{"_35":393,"_37":394,"_39":395,"_41":396,"_43":397},"2023-05-03-organizations","Organizations","2023-05-03T16:13:12Z","

You can now use an organization to collaborate with other people on Replicate.

Organizations let you share access to models, API tokens, billing, dashboards, and more. When you run models as the organization, it gets billed to your shared credit card instead of your personal account.

To get started, use the new account menu to create your organization:

\n $\"A$ ","May 3, 2023",{"_35":399,"_37":400,"_39":401,"_41":402,"_43":403},"2023-04-04-nodejs-client","Node.js client library","2023-04-04T17:16:05Z","

Replicate now has a client library for Node.js. You can use it to run models and everything else you can do with the HTTP API.

Install it from npm:

npm install replicate

Then, you can run predictions:

import Replicate from "replicate";\n \nconst replicate = new Replicate({\n  auth: process.env.REPLICATE_API_TOKEN,\n});\n \nconst model = "stability-ai/stable-diffusion:27b93a2413e7f36cd83da926f3656280b2931564ff050bf9575f1fdf9bcd7478";\nconst input = { prompt: "a 19th century portrait of a raccoon gentleman wearing a suit" };\nconst output = await replicate.run(model, { input });\n// ['https://replicate.delivery/pbxt/GtQb3Sgve42ZZyVnt8xjquFk9EX5LP0fF68NTIWlgBMUpguQA/out-0.png']

Follow our guide to running a model from Node.js or read the full documentation on GitHub.

","April 4, 2023",{"_35":405,"_37":406,"_39":407,"_41":408,"_43":409},"2023-03-21-more-useful-metadata-from-the-model-api","More useful metadata from the model API","2023-03-21T20:16:26Z","

The "get a model" API operation now returns more metadata about the model:

run_count: an integer indicating how many times the model has been run.
default_example: a prediction object created with this model, and selected by the model owner as an example of the model's inputs and outputs.
cover_image_url: an HTTPS URL string for an image file. This is an image uploaded by the model author, or an output file or input file from the model's default example prediction.

Here's an example using cURL and jq to get the Salesforce blip-2 model as a JSON object and pluck out some of its properties:

curl -s \\\n  -H "Authorization: Token $REPLICATE_API_TOKEN" \\\n  -H 'Content-Type: application/json' \\\n  "https://api.replicate.com/v1/models/salesforce/blip-2" \\\n  | jq "{owner, name, run_count, cover_image_url, default_example}"

Here's what the response looks like:

{\n  "owner": "salesforce",\n  "name": "blip-2",\n  "run_count": 270306,\n  "cover_image_url": "https://replicate.delivery/pbxt/IJEPmgAlL2zNBNDoRRKFegTEcxnlRhoQxlNjPHSZEy0pSIKn/gg_bridge.jpeg",\n  "default_example": {\n    "completed_at": "2023-02-13T22:26:49.396028Z",\n    "created_at": "2023-02-13T22:26:48.385476Z",\n    "error": null,\n    "id": "uhd4lhedtvdlbnm2cyhzx65zpe",\n    "input": {\n      "image": "https://replicate.delivery/pbxt/IJEPmgAlL2zNBNDoRRKFegTEcxnlRhoQxlNjPHSZEy0pSIKn/gg_bridge.jpeg",\n      "caption": false,\n      "question": "what body of water does this bridge cross?",\n      "temperature": 1\n    },\n    "logs": "...",\n    "metrics": {\n      "predict_time": 0.949567\n    },\n    "output": "san francisco bay",\n    "started_at": "2023-02-13T22:26:48.446461Z",\n    "status": "succeeded",\n    "version": "4b32258c42e9efd4288bb9910bc532a69727f9acd26aa08e175713a0a857a608",\n  }\n}

See the "get a model" API docs for more details.

","March 21, 2023",{"_35":411,"_37":412,"_39":413,"_41":414,"_43":415},"2023-03-20-get-model-input-and-output-schemas-via-the-api","Get model input and output schemas via the API","2023-03-20T21:07:49Z","

Every model on Replicate describes its inputs and outputs with OpenAPI Schema Objects in the openapi_schema property. This is a structured JSON object that includes the name, description, type, and allowed values for each input or output parameter.

Today we've improved our API reference documentation to clarify how to get a model's input and output schema.

See the updated docs at https://replicate.com/docs/reference/http#models.versions.get

Here's an example of how to get the input schema for Stable Diffusion using cURL:

curl -s \\\n  -H "Authorization: Token $REPLICATE_API_TOKEN" \\\n  -H 'Content-Type: application/json' \\\n  "https://api.replicate.com/v1/models/stability-ai/stable-diffusion/versions/db21e45d3f7023abc2a46ee38a23973f6dce16bb082a930b0c49861f96d1e5bf" | jq ".openapi_schema.components.schemas.Input.properties"

Using this command, we can see all the inputs to Stable Diffusion, including their types, description, min and max values, etc:

{\n  "seed": {\n    "type": "integer",\n    "title": "Seed",\n    "x-order": 7,\n    "description": "Random seed. Leave blank to randomize the seed"\n  },\n  "prompt": {\n    "type": "string",\n    "title": "Prompt",\n    "default": "a vision of paradise. unreal engine",\n    "x-order": 0,\n    "description": "Input prompt"\n  },\n  "scheduler": {\n    "allOf": [\n      {\n        "$ref": "#/components/schemas/scheduler"\n      }\n    ],\n    "default": "DPMSolverMultistep",\n    "x-order": 6,\n    "description": "Choose a scheduler."\n  },\n  "num_outputs": {\n    "type": "integer",\n    "title": "Num Outputs",\n    "default": 1,\n    "maximum": 4,\n    "minimum": 1,\n    "x-order": 3,\n    "description": "Number of images to output."\n  },\n  "guidance_scale": {\n    "type": "number",\n    "title": "Guidance Scale",\n    "default": 7.5,\n    "maximum": 20,\n    "minimum": 1,\n    "x-order": 5,\n    "description": "Scale for classifier-free guidance"\n  },\n  "negative_prompt": {\n    "type": "string",\n    "title": "Negative Prompt",\n    "x-order": 2,\n    "description": "Specify things to not see in the output"\n  },\n  "image_dimensions": {\n    "allOf": [\n      {\n        "$ref": "#/components/schemas/image_dimensions"\n      }\n    ],\n    "default": "768x768",\n    "x-order": 1,\n    "description": "pixel dimensions of output image"\n  },\n  "num_inference_steps": {\n    "type": "integer",\n    "title": "Num Inference Steps",\n    "default": 50,\n    "maximum": 500,\n    "minimum": 1,\n    "x-order": 4,\n    "description": "Number of denoising steps"\n  }\n}

And here's a command to the get the output schema:

curl -s \\\n  -H "Authorization: Token $REPLICATE_API_TOKEN" \\\n  -H 'Content-Type: application/json' \\\n  "https://api.replicate.com/v1/models/stability-ai/stable-diffusion/versions/db21e45d3f7023abc2a46ee38a23973f6dce16bb082a930b0c49861f96d1e5bf" | jq ".openapi_schema.components.schemas.Output"

From this command, we can see that Stable Diffusion output format is a list of URL strings:

{\n  "type": "array",\n  "items": {\n    "type": "string",\n    "format": "uri"\n  },\n  "title": "Output"\n}

","March 20, 2023",{"_35":417,"_37":418,"_39":419,"_41":420,"_43":421},"2023-02-16-openapi-schema","OpenAPI schema for the public API","2023-02-16T20:23:34Z","

OpenAPI (formerly known as Swagger) is a specification for describing HTTP APIs with structured data, including the available endpoints, their HTTP methods, expected request and response formats, and other metadata.

You can use this schema as a reference document, or to auto-generate your own API client code, or to mock and test your existing use of Replicate's API.

All of the documentation you see on the HTTP API reference page is also available as a structured OpenAPI JSON schema.

Download the schema at api.replicate.com/openapi.json

","February 16, 2023",{"_35":423,"_37":424,"_39":425,"_41":426,"_43":427},"2023-02-15-see-more-models","See more models","2023-02-15T17:38:27Z","

You can now browse through all the models on Replicate. Check them out on the Explore page!

\n $\"Screenshot$ ","February 15, 2023",{"_35":429,"_37":430,"_39":431,"_41":432,"_43":433},"2023-02-10-improved-webhook-events-and-event-filtering","Improved webhook events and event filtering","2023-02-10T22:10:31Z","

When you create a prediction with the API, you can provide a webhook URL for us to call when your prediction is complete.

Starting today, we now send more webhook events at different stages of the prediction lifecycle. We send requests to your webhook URL whenever there are new logs, new outputs, or the prediction has finished.

You can change which events trigger webhook requests by specifying webhook_events_filter in the prediction request.

start: Emitted immediately on prediction start. This event is always sent.
output: Emitted each time a prediction generates an output (note that predictions can generate multiple outputs).
logs: Emitted each time log output is generated by a prediction.
completed: Emitted when the prediction reaches a terminal state (succeeded/canceled/failed). This event is always sent.

For example, if you only wanted requests to be sent at the start and end of the prediction, you would provide:

{\n  "version": "5c7d5dc6dd8bf75c1acaa8565735e7986bc5b66206b55cca93cb72c9bf15ccaa",\n  "input": {\n    "text": "Alice"\n  },\n  "webhook": "https://example.com/my-webhook",\n  "webhook_events_filter": ["start", "completed"]\n}

Requests for event types output and logs will be sent at most once every 500ms.

Requests for event types start and completed will always be sent.

If you're using the old webhook_completed, you'll still get the same webhooks as before, but we recommend updating to use the new webhook and webhook_completed properties.

Docs: https://replicate.com/docs/reference/http#create-prediction--webhook

","February 10, 2023",{"_35":435,"_37":436,"_39":437,"_41":438,"_43":439},"2023-01-19-python-example-code-improvements","Python example code improvements","2023-01-19T18:33:16Z","

We've made it even easier to start building with Replicate's API. When you click on the API tab for a model, the Python example code now has everything you need to run a prediction, including code for all of the model's inputs and outputs. These new code snippets include documentation and defaults for each input, so you can focus on coding, with less context switching between the API docs and your editor.

To learn more about how to get started, check out our "Run a model from Python" guide.

\n $\"Python$ ","January 19, 2023",{"_35":441,"_37":442,"_39":443,"_41":444,"_43":445},"2023-01-16-cancel-long-running-predictions","Cancel long running predictions","2023-01-16T11:49:34Z","

Have you ever kicked off a prediction, then, after thinking about it, realized you got one of the settings wrong or wanted to tweak the prompt? Well, now you can cancel that prediction, even if you've navigated away from the page you were on or made it through the API. On the website, go to your dashboard, find the running prediction, and you'll now see it live-updating with a handy "Cancel" button.

\n $\"A$ ","January 16, 2023",{"_35":447,"_37":448,"_39":449,"_41":450,"_43":451},"2023-01-10-brew-install-cog","brew install cog","2023-01-10T00:34:45Z","

🍏 Hey macOS users! There's now a Homebrew formula for Cog. Use brew install cog to install it, and brew upgrade cog to upgrade to the latest version. See https://github.com/replicate/cog#install

","January 10, 2023",{"_35":453,"_37":454,"_39":455,"_41":456,"_43":457},"2023-01-09-dreambooth-support-for-img2img","Dreambooth support for img2img","2023-01-09T20:51:31Z","

We've added img2img support to models created with our DreamBooth API.

This means you can optionally send both a prompt and initial image to generate new images (in addition to the other parameters specified in your DreamBooth model's API page).

Input:

prompt: photo of zeke playing guitar on stage at concert

image: https://www.pexels.com/photo/man-playing-red-and-black-electric-guitar-on-stage-167382/

$\"base\"/$

Output:

$\"zeke\"/$

To get started building and pushing your own DreamBooth model, check out the blog post.

","January 9, 2023",{"_35":459,"_37":460,"_39":461,"_41":462,"_43":463},"2023-01-06-delete-predictions-from-the-web","Delete predictions from the web","2023-01-06T03:09:24Z","

You can now manually delete a prediction on the website. You'll find a "Delete" button on the prediction detail page, e.g. https://replicate.com/p/{prediction_id}. Clicking this link will completely remove the prediction from the site, including any output data and output files associated with it.

\n $\"delete-prediction\"$ ","January 6, 2023",{"_35":465,"_37":466,"_39":467,"_41":468,"_43":469},"2023-01-04-api-prediction-data-no-longer-stored","API prediction data no longer stored","2023-01-04T17:50:26Z","

By popular request, we no longer store data for predictions made using the API.

User data is automatically removed from predictions an hour after they finish. The prediction itself is not deleted, but the input and output data for the prediction is removed. This only applies to predictions created with the API, but not predictions created on the website.

This is enabled for new accounts starting today, but we know that some users may be relying on prediction data to exist for more than an hour, so we've not enabled this for any existing accounts.

","January 4, 2023",{"_35":471,"_37":472,"_39":473,"_41":474,"_43":475},"2022-12-06-a-proper-changelog","A proper changelog","2022-12-06T00:00:00Z","

We now have a changelog for product updates at replicate.com/changelog. We used to use a single tweet thread as our makeshift changelog, but decided it was time to make something a bit more flexible. Stay tuned for more frequent updates here!

","December 6, 2022",{"_35":477,"_37":478,"_39":479,"_41":480,"_43":481},"2022-12-02-stable-diffusion-has-release-notes","Stable diffusion has release notes","2022-12-02T00:00:00Z","

Stable Diffusion now has release notes so you can see what's changed: replicate.com/stability-ai/stable-diffusion/versions. It only works on Stable Diffusion at the moment. Coming soon to all models so you can set release notes on your own models and see what has changed on other people's models.

","December 2, 2022",{"_35":483,"_37":484,"_39":485,"_41":486,"_43":487},"2022-11-04-higher-rate-limits","Higher rate limits","2022-11-04T00:00:00Z","

We've also increased our default rate limits. You can create 10 predictions a second, bursting up to 600 predictions a second. https://replicate.com/docs/reference/http#rate-limits

We can support higher rates too – just email us: team@replicate.com

","November 4, 2022",{"_35":489,"_37":490,"_39":485,"_41":491,"_43":487},"2022-11-04-infrastructure-improvements","Infrastructure improvements","

Over the past few weeks, we've made some major improvements to our infrastructure to make it more reliable and perform better. Nothing's changed from your point of view, but you'll be seeing faster response times! 🚀

",{"_35":493,"_37":494,"_39":495,"_41":496,"_43":497},"2022-10-19-run-your-own-models-on-nvidia-a100-gpus","Run your own models on Nvidia A100 GPUs","2022-10-19T00:00:00Z","

You can now run your own models on Nvidia A100s. Click the settings tab on your model and select the hardware option to upgrade. 🚀

","October 19, 2022",{"_35":499,"_37":500,"_39":501,"_41":502,"_43":503},"2022-10-17-set-a-monthly-spend-limit","Set a monthly spend limit","2022-10-17T00:00:00Z","

You can now set a monthly spend limit on your account to avoid getting a surprising bill. 🦆

To set a limit, visit https://replicate.com/account#limits

","October 17, 2022",{"_35":505,"_37":506,"_39":507,"_41":508,"_43":509},"2022-09-09-webhook-support-in-predictions-api","Webhook support in Predictions API","2022-09-09T00:00:00Z","

🪝 Our API now supports webhooks, as an alternative to polling. Specify your webhook URL when creating a prediction and we'll POST to that URL when your prediction has completed! See the API docs here: https://replicate.com/docs/reference/http#create-prediction

","September 9, 2022",{"_35":511,"_37":512,"_39":513,"_41":514,"_43":515},"2022-06-16-introducing-model-collections","Introducing model collections","2022-06-16T00:00:00Z","

"We've started curating collections of models that perform similar tasks. First up is an assortment of ✨style transfer✨ models that take a content image and a style reference to produce a new image, like this starry night cat. https://replicate.com/collections/style-transfer

","June 16, 2022",{"_35":517,"_37":518,"_39":519,"_41":520,"_43":521},"2022-03-28-scrubbing-support-for-progressive-outputs","Scrubbing support for progressive outputs","2022-03-28T00:00:00Z","

"When you run a model that changes over time, you can scrub back and forth to see the previous output. We've now added that scrubber to predictions in the example gallery, so you can see how they morphed into being: https://replicate.com/pixray/text2image/examples

","March 28, 2022","routes/changelog/detail",{"_35":100},"content/changelog/2024-12-13-gpu-memory-monitoring","actionData","errors"]