mbukerepo / photomaker

PhotoMaker: Customizing Realistic Human Photos via Stacked ID Embedding

Cold

Public
4.8K runs
L40S
Paper
License

Run with an API

Playground API Examples README Versions

Input

first_image

*file

The input image, for example a photo of your face.

second_image

file

Additional input image (optional)

third_image

file

Additional input image (optional)

fourth_image

file

Additional input image (optional)

prompt

string

Shift + Return to add a new line

sci-fi, closeup portrait photo of a man img wearing the sunglasses in Iron man suit, face, slim body, high quality, film grainsci-fi, closeup portrait photo of a man img wearing the sunglasses in Iron man suit, face, slim body, high quality, film grain

Prompt. Example: 'a photo of a man/woman img'. The phrase 'img' is the trigger word.

Default: "A photo of a person img"

style_name

string

Style template. The style template will add a style-specific prompt and negative prompt to the user's prompt.

Default: "Photographic (Default)"

negative_prompt

string

Shift + Return to add a new line

(asymmetry, worst quality, low quality, illustration, 3d, 2d, painting, cartoons, sketch), open mouth(asymmetry, worst quality, low quality, illustration, 3d, 2d, painting, cartoons, sketch), open mouth

Negative Prompt. The negative prompt should NOT contain the trigger word.

Default: "nsfw, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry"

num_steps

integer

(minimum: 1, maximum: 100)

Number of sample steps

Default: 20

style_strength_ratio

number

(minimum: 15, maximum: 50)

Style strength (%)

Default: 20

num_outputs

integer

(minimum: 1, maximum: 4)

Number of output images

Default: 1

guidance_scale

number

(minimum: 1, maximum: 10)

Guidance scale. A guidance scale of 1 corresponds to doing no classifier free guidance.

Default: 5

seed

integer

(minimum: 0, maximum: 2147483647)

Seed. Leave blank to use a random number

disable_safety_checker

boolean

Disable safety checker for generated images.

Default: false

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

npx create-replicate --model=mbukerepo/photomaker

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 mbukerepo/photomaker using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

const output = await replicate.run(
  "mbukerepo/photomaker:2cef1f7ff2e6fa2ba24569ddc467fe8e706db0f96e4663f4b5df213469c8e3a2",
  {
    input: {
      prompt: "sci-fi, closeup portrait photo of a man img wearing the sunglasses in Iron man suit, face, slim body, high quality, film grain",
      num_steps: 20,
      style_name: "Photographic (Default)",
      num_outputs: 1,
      guidance_scale: 5,
      negative_prompt: "(asymmetry, worst quality, low quality, illustration, 3d, 2d, painting, cartoons, sketch), open mouth",
      style_strength_ratio: 20
    }
  }
);

// 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 mbukerepo/photomaker using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

output = replicate.run(
    "mbukerepo/photomaker:2cef1f7ff2e6fa2ba24569ddc467fe8e706db0f96e4663f4b5df213469c8e3a2",
    input={
        "prompt": "sci-fi, closeup portrait photo of a man img wearing the sunglasses in Iron man suit, face, slim body, high quality, film grain",
        "num_steps": 20,
        "style_name": "Photographic (Default)",
        "num_outputs": 1,
        "guidance_scale": 5,
        "negative_prompt": "(asymmetry, worst quality, low quality, illustration, 3d, 2d, painting, cartoons, sketch), open mouth",
        "style_strength_ratio": 20
    }
)
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 mbukerepo/photomaker 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": "mbukerepo/photomaker:2cef1f7ff2e6fa2ba24569ddc467fe8e706db0f96e4663f4b5df213469c8e3a2",
    "input": {
      "prompt": "sci-fi, closeup portrait photo of a man img wearing the sunglasses in Iron man suit, face, slim body, high quality, film grain",
      "num_steps": 20,
      "style_name": "Photographic (Default)",
      "num_outputs": 1,
      "guidance_scale": 5,
      "negative_prompt": "(asymmetry, worst quality, low quality, illustration, 3d, 2d, painting, cartoons, sketch), open mouth",
      "style_strength_ratio": 20
    }
  }' \
  https://api.replicate.com/v1/predictions

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

Output

{
  "completed_at": "2024-01-16T19:44:41.073782Z",
  "created_at": "2024-01-16T19:43:34.988266Z",
  "data_removed": false,
  "error": null,
  "id": "cnefqklbpmgqnbsgslcytb6fwy",
  "input": {
    "image": "https://replicate.delivery/pbxt/KEoNTjDyiTvEKYmQF6d26yqIl7smGBQvrEKXva1qZg0DzV3j/newton_0.jpg",
    "prompt": "sci-fi, closeup portrait photo of a man img wearing the sunglasses in Iron man suit, face, slim body, high quality, film grain",
    "num_outputs": 1,
    "negative_prompt": "(asymmetry, worst quality, low quality, illustration, 3d, 2d, painting, cartoons, sketch), open mouth",
    "num_inference_steps": 40
  },
  "logs": "2024-01-16 19:44:30,358 INFO predict Using seed: 318775383\n  0%|          | 0/40 [00:00<?, ?it/s]\n  2%|▎         | 1/40 [00:00<00:09,  4.33it/s]\n  5%|▌         | 2/40 [00:00<00:08,  4.51it/s]\n  8%|▊         | 3/40 [00:00<00:08,  4.57it/s]\n 10%|█         | 4/40 [00:00<00:07,  4.60it/s]\n 12%|█▎        | 5/40 [00:01<00:07,  4.61it/s]\n 15%|█▌        | 6/40 [00:01<00:07,  4.62it/s]\n 18%|█▊        | 7/40 [00:01<00:07,  4.62it/s]\n 20%|██        | 8/40 [00:01<00:06,  4.62it/s]\n 22%|██▎       | 9/40 [00:01<00:06,  4.62it/s]\n 25%|██▌       | 10/40 [00:02<00:06,  4.63it/s]\n 28%|██▊       | 11/40 [00:02<00:06,  4.61it/s]\n 30%|███       | 12/40 [00:02<00:06,  4.62it/s]\n 32%|███▎      | 13/40 [00:02<00:05,  4.63it/s]\n 35%|███▌      | 14/40 [00:03<00:05,  4.63it/s]\n 38%|███▊      | 15/40 [00:03<00:05,  4.63it/s]\n 40%|████      | 16/40 [00:03<00:05,  4.64it/s]\n 42%|████▎     | 17/40 [00:03<00:04,  4.64it/s]\n 45%|████▌     | 18/40 [00:03<00:04,  4.64it/s]\n 48%|████▊     | 19/40 [00:04<00:04,  4.63it/s]\n 50%|█████     | 20/40 [00:04<00:04,  4.63it/s]\n 52%|█████▎    | 21/40 [00:04<00:04,  4.63it/s]\n 55%|█████▌    | 22/40 [00:04<00:03,  4.63it/s]\n 57%|█████▊    | 23/40 [00:04<00:03,  4.63it/s]\n 60%|██████    | 24/40 [00:05<00:03,  4.62it/s]\n 62%|██████▎   | 25/40 [00:05<00:03,  4.62it/s]\n 65%|██████▌   | 26/40 [00:05<00:03,  4.63it/s]\n 68%|██████▊   | 27/40 [00:05<00:02,  4.63it/s]\n 70%|███████   | 28/40 [00:06<00:02,  4.62it/s]\n 72%|███████▎  | 29/40 [00:06<00:02,  4.63it/s]\n 75%|███████▌  | 30/40 [00:06<00:02,  4.62it/s]\n 78%|███████▊  | 31/40 [00:06<00:01,  4.62it/s]\n 80%|████████  | 32/40 [00:06<00:01,  4.63it/s]\n 82%|████████▎ | 33/40 [00:07<00:01,  4.63it/s]\n 85%|████████▌ | 34/40 [00:07<00:01,  4.63it/s]\n 88%|████████▊ | 35/40 [00:07<00:01,  4.63it/s]\n 90%|█████████ | 36/40 [00:07<00:00,  4.63it/s]\n 92%|█████████▎| 37/40 [00:08<00:00,  4.63it/s]\n 95%|█████████▌| 38/40 [00:08<00:00,  4.63it/s]\n 98%|█████████▊| 39/40 [00:08<00:00,  4.62it/s]\n100%|██████████| 40/40 [00:08<00:00,  4.63it/s]\n100%|██████████| 40/40 [00:08<00:00,  4.62it/s]",
  "metrics": {
    "predict_time": 11.242848,
    "total_time": 66.085516
  },
  "output": [
    "https://replicate.delivery/pbxt/VpepbXCfqspATE8k9WMVdRqCr7VvEsJ89ImSFMQNPxRo6INSA/out-0.png"
  ],
  "started_at": "2024-01-16T19:44:29.830934Z",
  "status": "succeeded",
  "urls": {
    "get": "https://api.replicate.com/v1/predictions/cnefqklbpmgqnbsgslcytb6fwy",
    "cancel": "https://api.replicate.com/v1/predictions/cnefqklbpmgqnbsgslcytb6fwy/cancel"
  },
  "version": "3be42163bbfb293a1f07af4d3db64f50068cbe8b5280d10d5944c0e2afae26d5"
}

Generated in

11.3 seconds

Tweak itReport

2024-01-16 19:44:30,358 INFO predict Using seed: 318775383
  0%|          | 0/40 [00:00<?, ?it/s]
  2%|▎         | 1/40 [00:00<00:09,  4.33it/s]
  5%|▌         | 2/40 [00:00<00:08,  4.51it/s]
  8%|▊         | 3/40 [00:00<00:08,  4.57it/s]
 10%|█         | 4/40 [00:00<00:07,  4.60it/s]
 12%|█▎        | 5/40 [00:01<00:07,  4.61it/s]
 15%|█▌        | 6/40 [00:01<00:07,  4.62it/s]
 18%|█▊        | 7/40 [00:01<00:07,  4.62it/s]
 20%|██        | 8/40 [00:01<00:06,  4.62it/s]
 22%|██▎       | 9/40 [00:01<00:06,  4.62it/s]
 25%|██▌       | 10/40 [00:02<00:06,  4.63it/s]
 28%|██▊       | 11/40 [00:02<00:06,  4.61it/s]
 30%|███       | 12/40 [00:02<00:06,  4.62it/s]
 32%|███▎      | 13/40 [00:02<00:05,  4.63it/s]
 35%|███▌      | 14/40 [00:03<00:05,  4.63it/s]
 38%|███▊      | 15/40 [00:03<00:05,  4.63it/s]
 40%|████      | 16/40 [00:03<00:05,  4.64it/s]
 42%|████▎     | 17/40 [00:03<00:04,  4.64it/s]
 45%|████▌     | 18/40 [00:03<00:04,  4.64it/s]
 48%|████▊     | 19/40 [00:04<00:04,  4.63it/s]
 50%|█████     | 20/40 [00:04<00:04,  4.63it/s]
 52%|█████▎    | 21/40 [00:04<00:04,  4.63it/s]
 55%|█████▌    | 22/40 [00:04<00:03,  4.63it/s]
 57%|█████▊    | 23/40 [00:04<00:03,  4.63it/s]
 60%|██████    | 24/40 [00:05<00:03,  4.62it/s]
 62%|██████▎   | 25/40 [00:05<00:03,  4.62it/s]
 65%|██████▌   | 26/40 [00:05<00:03,  4.63it/s]
 68%|██████▊   | 27/40 [00:05<00:02,  4.63it/s]
 70%|███████   | 28/40 [00:06<00:02,  4.62it/s]
 72%|███████▎  | 29/40 [00:06<00:02,  4.63it/s]
 75%|███████▌  | 30/40 [00:06<00:02,  4.62it/s]
 78%|███████▊  | 31/40 [00:06<00:01,  4.62it/s]
 80%|████████  | 32/40 [00:06<00:01,  4.63it/s]
 82%|████████▎ | 33/40 [00:07<00:01,  4.63it/s]
 85%|████████▌ | 34/40 [00:07<00:01,  4.63it/s]
 88%|████████▊ | 35/40 [00:07<00:01,  4.63it/s]
 90%|█████████ | 36/40 [00:07<00:00,  4.63it/s]
 92%|█████████▎| 37/40 [00:08<00:00,  4.63it/s]
 95%|█████████▌| 38/40 [00:08<00:00,  4.63it/s]
 98%|█████████▊| 39/40 [00:08<00:00,  4.62it/s]
100%|██████████| 40/40 [00:08<00:00,  4.63it/s]
100%|██████████| 40/40 [00:08<00:00,  4.62it/s]

This output was created using a different version of the model, mbukerepo/photomaker:3be42163.

Examples

View more examples

Run time and cost

This model costs approximately $0.035 to run on Replicate, or 28 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 37 seconds.

Readme

PhotoMaker

Customizing Realistic Human Photos via Stacked ID Embedding .

Usage

Users can input one or a few face photos, along with a text prompt, to receive a customized photo or painting within seconds (no training required!). Additionally, this model can be adapted to any base model based on SDXL or used in conjunction with other LoRA modules.

Realistic results

image/jpeg

Stylization results

image/jpeg

More results can be found in our project page

Model Details

It mainly contains two parts corresponding to two keys in loaded state dict:

id_encoder includes finetuned OpenCLIP-ViT-H-14 and a few fuse layers.
lora_weights applies to all attention layers in the UNet, and the rank is set to 64.