fofr/sd3-explorer – Run with an API on Replicate

fofr / sd3-explorer

A model for experimenting with all the SD3 settings. Non-commercial use only, unless you have a Stability AI Self Hosted License.

Cold

Public
32.2K runs
L40S
GitHub
Paper
License

Run with an API

Playground API Examples README Versions

Input

prompt

string

Shift + Return to add a new line

a man and woman are standing together against a backdrop, the backdrop is divided equally in half down the middle, left side is red, right side is gold, the woman is wearing a t-shirt with a yoda motif, she has a long skirt with birds on it, the man is wearing a three piece purple suit, he has spiky blue haira man and woman are standing together against a backdrop, the backdrop is divided equally in half down the middle, left side is red, right side is gold, the woman is wearing a t-shirt with a yoda motif, she has a long skirt with birds on it, the man is wearing a three piece purple suit, he has spiky blue hair

This prompt is ignored when using the triple prompt mode. See below.

Default: ""

model

string

Pick whether to use T5-XXL in fp16, fp8 or not at all. We recommend fp16 for this model as it has the best image quality. When running locally we recommend fp8 for lower memory usage. We've included all versions here for exploration.

Default: "sd3_medium_incl_clips_t5xxlfp16.safetensors"

width

integer

The width of the image (best output at ~1 megapixel. Resolution must be divisible by 64)

Default: 1024

height

integer

The height of the image (best output at ~1 megapixel. Resolution must be divisible by 64)

Default: 1024

steps

integer

The number of steps to run the model for (more steps = better image but slower generation. Best results for this model are around 26 to 36 steps.)

Default: 28

sampler

string

The sampler to use (used to manage noise)

Default: "dpmpp_2m"

scheduler

string

The scheduler to use (used to manage noise; do not use karras)

Default: "sgm_uniform"

shift

number

(minimum: 0, maximum: 20)

The timestep scheduling shift; shift values higher than 1.0 are better at managing noise in higher resolutions. Try values 6.0 and 2.0 to experiment with effects.

Default: 3

guidance_scale

number

(minimum: 0, maximum: 20)

The guidance scale tells the model how similar the output should be to the prompt. (Recommend between 3.5 and 4.5; if images look 'burnt,' lower the value.)

Default: 3.5

number_of_images

integer

(minimum: 1, maximum: 10)

The number of images to generate

Default: 1

use_triple_prompt

boolean

Default: false

triple_prompt_clip_g

string

Shift + Return to add a new line

The prompt that will be passed to just the CLIP-G model.

Default: ""

triple_prompt_clip_l

string

Shift + Return to add a new line

The prompt that will be passed to just the CLIP-L model.

Default: ""

triple_prompt_t5

string

Shift + Return to add a new line

The prompt that will be passed to just the T5-XXL model.

Default: ""

triple_prompt_empty_padding

boolean

Whether to add padding for empty prompts. Useful if you only want to pass a prompt to one or two of the three text encoders. Has no effect when all prompts are filled. Disable this for interesting effects.

Default: true

negative_prompt

string

Shift + Return to add a new line

Negative prompts do not really work in SD3. This will simply cause your output image to vary in unpredictable ways.

Default: ""

negative_conditioning_end

number

(minimum: 0, maximum: 1)

When the negative conditioning should stop being applied. By default it is disabled. If you want to try a negative prompt, start with a value of 0.1

Default: 0

output_format

string

Format of the output images

Default: "webp"

output_quality

integer

(minimum: 0, maximum: 100)

Quality of the output images, from 0 to 100. 100 is best quality, 0 is lowest quality.

Default: 80

seed

integer

Set a seed for reproducibility. Random by default.

Run this model in Node.js with one line of code:

npx create-replicate --model=fofr/sd3-explorer

or set up a project from scratch

Install Replicate’s Node.js client library:

npm install replicate

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Import and set up the client:

import Replicate from "replicate";

const replicate = new Replicate({
  auth: process.env.REPLICATE_API_TOKEN,
});

Run fofr/sd3-explorer using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

const output = await replicate.run(
  "fofr/sd3-explorer:a9f4aebd943ad7db13de8e34debea359d5578d08f128e968f9a36c3e9b0148d4",
  {
    input: {
      model: "sd3_medium_incl_clips_t5xxlfp16.safetensors",
      shift: 3,
      steps: 28,
      width: 1024,
      height: 1024,
      prompt: "a man and woman are standing together against a backdrop, the backdrop is divided equally in half down the middle, left side is red, right side is gold, the woman is wearing a t-shirt with a yoda motif, she has a long skirt with birds on it, the man is wearing a three piece purple suit, he has spiky blue hair",
      sampler: "dpmpp_2m",
      scheduler: "sgm_uniform",
      output_format: "webp",
      guidance_scale: 4.5,
      output_quality: 80,
      negative_prompt: "",
      number_of_images: 1,
      triple_prompt_t5: "",
      use_triple_prompt: false,
      triple_prompt_clip_g: "",
      triple_prompt_clip_l: "",
      negative_conditioning_end: 0,
      triple_prompt_empty_padding: true
    }
  }
);

// To access the file URL:
console.log(output[0].url()); //=> "http://example.com"

// To write the file to disk:
fs.writeFile("my-image.png", output[0]);

To learn more, take a look at the guide on getting started with Node.js.

Install Replicate’s Python client library:

pip install replicate

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Import the client:

import replicate

Run fofr/sd3-explorer using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

output = replicate.run(
    "fofr/sd3-explorer:a9f4aebd943ad7db13de8e34debea359d5578d08f128e968f9a36c3e9b0148d4",
    input={
        "model": "sd3_medium_incl_clips_t5xxlfp16.safetensors",
        "shift": 3,
        "steps": 28,
        "width": 1024,
        "height": 1024,
        "prompt": "a man and woman are standing together against a backdrop, the backdrop is divided equally in half down the middle, left side is red, right side is gold, the woman is wearing a t-shirt with a yoda motif, she has a long skirt with birds on it, the man is wearing a three piece purple suit, he has spiky blue hair",
        "sampler": "dpmpp_2m",
        "scheduler": "sgm_uniform",
        "output_format": "webp",
        "guidance_scale": 4.5,
        "output_quality": 80,
        "negative_prompt": "",
        "number_of_images": 1,
        "triple_prompt_t5": "",
        "use_triple_prompt": False,
        "triple_prompt_clip_g": "",
        "triple_prompt_clip_l": "",
        "negative_conditioning_end": 0,
        "triple_prompt_empty_padding": True
    }
)
print(output)

To learn more, take a look at the guide on getting started with Python.

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Run fofr/sd3-explorer using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

curl -s -X POST \
  -H "Authorization: Bearer $REPLICATE_API_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Prefer: wait" \
  -d $'{
    "version": "a9f4aebd943ad7db13de8e34debea359d5578d08f128e968f9a36c3e9b0148d4",
    "input": {
      "model": "sd3_medium_incl_clips_t5xxlfp16.safetensors",
      "shift": 3,
      "steps": 28,
      "width": 1024,
      "height": 1024,
      "prompt": "a man and woman are standing together against a backdrop, the backdrop is divided equally in half down the middle, left side is red, right side is gold, the woman is wearing a t-shirt with a yoda motif, she has a long skirt with birds on it, the man is wearing a three piece purple suit, he has spiky blue hair",
      "sampler": "dpmpp_2m",
      "scheduler": "sgm_uniform",
      "output_format": "webp",
      "guidance_scale": 4.5,
      "output_quality": 80,
      "negative_prompt": "",
      "number_of_images": 1,
      "triple_prompt_t5": "",
      "use_triple_prompt": false,
      "triple_prompt_clip_g": "",
      "triple_prompt_clip_l": "",
      "negative_conditioning_end": 0,
      "triple_prompt_empty_padding": true
    }
  }' \
  https://api.replicate.com/v1/predictions

To learn more, take a look at Replicate’s HTTP API reference docs.

Output

{
  "completed_at": "2024-06-18T11:59:23.602030Z",
  "created_at": "2024-06-18T11:59:14.919000Z",
  "data_removed": false,
  "error": null,
  "id": "z78h5v1dwxrgp0cg5cvrpgs2fc",
  "input": {
    "model": "sd3_medium_incl_clips_t5xxlfp16.safetensors",
    "shift": 3,
    "steps": 28,
    "width": 1024,
    "height": 1024,
    "prompt": "a man and woman are standing together against a backdrop, the backdrop is divided equally in half down the middle, left side is red, right side is gold, the woman is wearing a t-shirt with a yoda motif, she has a long skirt with birds on it, the man is wearing a three piece purple suit, he has spiky blue hair",
    "sampler": "dpmpp_2m",
    "scheduler": "sgm_uniform",
    "output_format": "webp",
    "guidance_scale": 4.5,
    "output_quality": 80,
    "negative_prompt": "",
    "number_of_images": 1,
    "triple_prompt_t5": "",
    "use_triple_prompt": false,
    "triple_prompt_clip_g": "",
    "triple_prompt_clip_l": "",
    "negative_conditioning_end": 0,
    "triple_prompt_empty_padding": true
  },
  "logs": "Random seed set to: 2483966773\nRunning workflow\ngot prompt\nExecuting node 6, title: CLIP Text Encode (Prompt), class type: CLIPTextEncode\nExecuting node 271, title: KSampler, class type: KSampler\n  0%|          | 0/28 [00:00<?, ?it/s]\n  4%|▎         | 1/28 [00:00<00:04,  5.89it/s]\n  7%|▋         | 2/28 [00:00<00:06,  3.93it/s]\n 11%|█         | 3/28 [00:00<00:06,  4.09it/s]\n 14%|█▍        | 4/28 [00:00<00:05,  4.15it/s]\n 18%|█▊        | 5/28 [00:01<00:05,  4.18it/s]\n 21%|██▏       | 6/28 [00:01<00:05,  4.22it/s]\n 25%|██▌       | 7/28 [00:01<00:04,  4.24it/s]\n 29%|██▊       | 8/28 [00:01<00:04,  4.25it/s]\n 32%|███▏      | 9/28 [00:02<00:04,  4.25it/s]\n 36%|███▌      | 10/28 [00:02<00:04,  4.26it/s]\n 39%|███▉      | 11/28 [00:02<00:03,  4.26it/s]\n 43%|████▎     | 12/28 [00:02<00:03,  4.26it/s]\n 46%|████▋     | 13/28 [00:03<00:03,  4.26it/s]\n 50%|█████     | 14/28 [00:03<00:03,  4.26it/s]\n 54%|█████▎    | 15/28 [00:03<00:03,  4.26it/s]\n 57%|█████▋    | 16/28 [00:03<00:02,  4.26it/s]\n 61%|██████    | 17/28 [00:03<00:02,  4.26it/s]\n 64%|██████▍   | 18/28 [00:04<00:02,  4.26it/s]\n 68%|██████▊   | 19/28 [00:04<00:02,  4.25it/s]\n 71%|███████▏  | 20/28 [00:04<00:01,  4.26it/s]\n 75%|███████▌  | 21/28 [00:04<00:01,  4.26it/s]\n 79%|███████▊  | 22/28 [00:05<00:01,  4.25it/s]\n 82%|████████▏ | 23/28 [00:05<00:01,  4.25it/s]\n 86%|████████▌ | 24/28 [00:05<00:00,  4.25it/s]\n 89%|████████▉ | 25/28 [00:05<00:00,  4.25it/s]\n 93%|█████████▎| 26/28 [00:06<00:00,  4.25it/s]\n 96%|█████████▋| 27/28 [00:06<00:00,  4.25it/s]\n100%|██████████| 28/28 [00:06<00:00,  4.24it/s]\n100%|██████████| 28/28 [00:06<00:00,  4.25it/s]\nExecuting node 231, title: VAE Decode, class type: VAEDecode\nExecuting node 273, title: Save Image, class type: SaveImage\nPrompt executed in 7.17 seconds\noutputs:  {'273': {'images': [{'filename': 'SD3_00001_.png', 'subfolder': '', 'type': 'output'}]}}\n====================================\nSD3_00001_.png",
  "metrics": {
    "predict_time": 8.645392249,
    "total_time": 8.68303
  },
  "output": [
    "https://replicate.delivery/pbxt/BnkJxF51oYZsBdGsgn6vGIkeQUO17GgTPloJCuM0LpQNR5fSA/SD3_00001_.webp"
  ],
  "started_at": "2024-06-18T11:59:14.956637Z",
  "status": "succeeded",
  "urls": {
    "get": "https://api.replicate.com/v1/predictions/z78h5v1dwxrgp0cg5cvrpgs2fc",
    "cancel": "https://api.replicate.com/v1/predictions/z78h5v1dwxrgp0cg5cvrpgs2fc/cancel"
  },
  "version": "7c48d3a1f9e16683c3f1e546056becfe31b963a872384458703f7aa6a40a2c69"
}

Generated in

8.7 seconds

Tweak it ShareReport

Random seed set to: 2483966773
Running workflow
got prompt
Executing node 6, title: CLIP Text Encode (Prompt), class type: CLIPTextEncode
Executing node 271, title: KSampler, class type: KSampler
  0%|          | 0/28 [00:00<?, ?it/s]
  4%|▎         | 1/28 [00:00<00:04,  5.89it/s]
  7%|▋         | 2/28 [00:00<00:06,  3.93it/s]
 11%|█         | 3/28 [00:00<00:06,  4.09it/s]
 14%|█▍        | 4/28 [00:00<00:05,  4.15it/s]
 18%|█▊        | 5/28 [00:01<00:05,  4.18it/s]
 21%|██▏       | 6/28 [00:01<00:05,  4.22it/s]
 25%|██▌       | 7/28 [00:01<00:04,  4.24it/s]
 29%|██▊       | 8/28 [00:01<00:04,  4.25it/s]
 32%|███▏      | 9/28 [00:02<00:04,  4.25it/s]
 36%|███▌      | 10/28 [00:02<00:04,  4.26it/s]
 39%|███▉      | 11/28 [00:02<00:03,  4.26it/s]
 43%|████▎     | 12/28 [00:02<00:03,  4.26it/s]
 46%|████▋     | 13/28 [00:03<00:03,  4.26it/s]
 50%|█████     | 14/28 [00:03<00:03,  4.26it/s]
 54%|█████▎    | 15/28 [00:03<00:03,  4.26it/s]
 57%|█████▋    | 16/28 [00:03<00:02,  4.26it/s]
 61%|██████    | 17/28 [00:03<00:02,  4.26it/s]
 64%|██████▍   | 18/28 [00:04<00:02,  4.26it/s]
 68%|██████▊   | 19/28 [00:04<00:02,  4.25it/s]
 71%|███████▏  | 20/28 [00:04<00:01,  4.26it/s]
 75%|███████▌  | 21/28 [00:04<00:01,  4.26it/s]
 79%|███████▊  | 22/28 [00:05<00:01,  4.25it/s]
 82%|████████▏ | 23/28 [00:05<00:01,  4.25it/s]
 86%|████████▌ | 24/28 [00:05<00:00,  4.25it/s]
 89%|████████▉ | 25/28 [00:05<00:00,  4.25it/s]
 93%|█████████▎| 26/28 [00:06<00:00,  4.25it/s]
 96%|█████████▋| 27/28 [00:06<00:00,  4.25it/s]
100%|██████████| 28/28 [00:06<00:00,  4.24it/s]
100%|██████████| 28/28 [00:06<00:00,  4.25it/s]
Executing node 231, title: VAE Decode, class type: VAEDecode
Executing node 273, title: Save Image, class type: SaveImage
Prompt executed in 7.17 seconds
outputs:  {'273': {'images': [{'filename': 'SD3_00001_.png', 'subfolder': '', 'type': 'output'}]}}
====================================
SD3_00001_.png

This output was created using a different version of the model, fofr/sd3-explorer:7c48d3a1.

Examples

View more examples

Run time and cost

This model costs approximately $0.0044 to run on Replicate, or 227 runs per $1, but this varies depending on your inputs. It is also open source and you can run it on your own computer with Docker.

This model runs on Nvidia L40S GPU hardware. Predictions typically complete within 5 seconds.

Readme

A model for exploring all the SD3 possibilities

This is for non-commercial use only and is intended for model exploration. If you want to use this model commercially, you should buy a Stability AI Self Hosted License:

https://stability.ai/license

Or use the official Stability AI model which is available for commercial use:

https://replicate.com/stability-ai/stable-diffusion-3