black-forest-labs / flux-pro-trainer

Train FLUX.1 [pro] and FLUX 1.1 [pro] Ultra. Upload images to create a custom finetune_id to use with the inference model

Warm

Public
1.8K runs
Priced by multiple properties
Commercial use
Paper
License

Iterate in playground

Run with an API

Playground API Examples README

Input

input_images

*file

A zip file containing the images that will be used for training. We recommend a minimum of 10 images. If you include captions, include them as one .txt file per image, e.g. my-photo.jpg should have a caption file named my-photo.txt. If you don't include captions, you can turn on captioning.

training_steps

integer

Number of training steps for the fine-tuning process.

Default: 300

trigger_word

string

Shift + Return to add a new line

The trigger word refers to the object, style or concept you are training on. Pick a string that isn't a real word, like TOK or something related to what's being trained, like CYBRPNK. The trigger word you specify here will be associated with all images during training. Then when you use your LoRA, you can include the trigger word in prompts to help activate the LoRA.

Default: "TOK"

mode

string

Mode for the fine-tuned model

Default: "general"

learning_rate

number

(minimum: 1e-7, maximum: 0.005)

Learning rate for fine-tuning. If not provided, defaults to 1e-5 for full fine-tuning and 1e-4 for lora fine-tuning

captioning

string

Whether to enable captioning during fine-tuning. Automatic turns on captioning if any images are missing caption files.

Default: "automatic"

priority

string

Priority of the fine-tuning process

Default: "quality"

finetune_type

string

Type of fine-tuning

Default: "full"

lora_rank

integer

Rank of the fine-tuned model

Default: 32

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

npx create-replicate --model=black-forest-labs/flux-pro-trainer

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 black-forest-labs/flux-pro-trainer using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

const input = {
  mode: "general",
  priority: "quality",
  lora_rank: 32,
  captioning: "captioning-disabled",
  input_images: "https://replicate.delivery/pbxt/MvDQOp1uNVTNjYxI9wnvyqXNMumlUzQdBFFPvkqNyIkkG1Rn/ABSRTBNGS.zip",
  trigger_word: "ABSRTBNGS",
  finetune_type: "full",
  training_steps: 300
};

const output = await replicate.run("black-forest-labs/flux-pro-trainer", { input });

console.log(output);

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 black-forest-labs/flux-pro-trainer using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

output = replicate.run(
    "black-forest-labs/flux-pro-trainer",
    input={
        "mode": "general",
        "priority": "quality",
        "lora_rank": 32,
        "captioning": "captioning-disabled",
        "input_images": "https://replicate.delivery/pbxt/MvDQOp1uNVTNjYxI9wnvyqXNMumlUzQdBFFPvkqNyIkkG1Rn/ABSRTBNGS.zip",
        "trigger_word": "ABSRTBNGS",
        "finetune_type": "full",
        "training_steps": 300
    }
)
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 black-forest-labs/flux-pro-trainer 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 $'{
    "input": {
      "mode": "general",
      "priority": "quality",
      "lora_rank": 32,
      "captioning": "captioning-disabled",
      "input_images": "https://replicate.delivery/pbxt/MvDQOp1uNVTNjYxI9wnvyqXNMumlUzQdBFFPvkqNyIkkG1Rn/ABSRTBNGS.zip",
      "trigger_word": "ABSRTBNGS",
      "finetune_type": "full",
      "training_steps": 300
    }
  }' \
  https://api.replicate.com/v1/models/black-forest-labs/flux-pro-trainer/predictions

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

Output

f07345cb-4d18-4461-aa6f-179566153d28

Generated in

11 minutes 20 seconds

Tweak it Report View full prediction

Pricing

Model pricing for black-forest-labs/flux-pro-trainer. Looking for volume pricing? Get in touch.

Official models are always on, maintained, and have predictable pricing. Learn more.

Check out our docs for more information about how pricing works on Replicate.

training step is ≥0 and <150

per run

or 50 runs for $100

training step is ≥150 and ≤500

per run

or 25 runs for $100

training step is >500

per run

or around 16 runs for $100

Readme

FLUX 1.1 Pro Ultra - Finetuner (Trainer)

This is the training model for creating custom finetunes compatible with both the FLUX 1.1 [pro] Ultra and FLUX.1 [pro] architectures. Use this model to teach FLUX new concepts, styles, or subjects based on your own images. The training process generates a unique finetune_id representing your custom model.

How it Works

Prepare Your Data: Gather a set of images (typically 5-20) that clearly represent the concept, style, or subject you want the model to learn. Ensure images are well-lit and focused on the subject.
Upload Data: Upload your images as a .zip or .tar.gz archive to the input_images field.
Define Your Concept (Optional but Recommended): Use the caption_prefix input to provide a unique identifier or trigger word for your concept (e.g., “a photo of sks style”). This helps during inference.
Start Training: Run the model. The training process will analyze your images and create a specialized finetune.
Get Your Finetune ID: Upon successful completion, the output will be your unique finetune_id. Keep this ID safe!

Using Your Finetune for Inference

Once you have your finetune_id, you can use it to generate images with your custom style or concept using the corresponding inference model:

On the inference page, give your finetune_id along with your text prompt. You can also adjust the finetune_strength to control how strongly your custom concept influences the generated image.

License

By using the trainer through Replicate you agree to the Black Forest Labs API agreement and the Black Forest Labs Terms of Service.