Collections

Flux fine-tunes

FLUX.1 is a family of text-to-image models released by Black Forest Labs. The FLUX.1 models set a new standard for open-weights image models: they can generate realistic hands, legible text, and even the strangely hard task of “funny memes.” The models in this collection are fine-tunes created on Replicate using the FLUX.1 Dev LoRA Trainer.

You can create your own fine-tuned Flux model using your own training images. Check out replicate.com/blog/fine-tune-flux to get started.

Frequently asked questions

What is the Use FLUX fine-tunes collection?

This collection includes models created by fine-tuning black-forest-labs/flux-1.1-pro from Black Forest Labs. These models let you generate images in a specific style, subject, or character that goes beyond what the base model can do.
They’re great if you want something more personal, branded, or niche than a general-purpose model.

How do I choose a fine-tuned model?

Each fine-tuned model has its own unique concept—like a particular character, art style, or visual theme.
Browse the collection, read the model descriptions, and note any trigger words they use (e.g., MYCHAR). Run a few sample prompts to see which fine-tune best fits your creative direction.

When should I make my own fine-tune instead of using an existing one?

You’ll want to train your own FLUX fine-tune if:

  • You need a very specific character, product, or style that doesn’t exist yet.
  • You want consistent results for commercial or branded work.
  • You need control over the trigger word and training data.
    Existing fine-tunes are great for testing ideas, but your own gives you full control.

What’s involved in fine-tuning FLUX on Replicate?

Fine-tuning typically involves:

  • Providing a small set of images (as few as two, though 10+ gives better results).
  • Uploading them to the fine-tuning interface or API.
  • Choosing a trigger word to activate the concept in prompts.
  • Training for a few minutes.
    Once done, your new model appears in your account and can be run like any other FLUX model.

What kind of inputs and outputs do these models use?

  • Inputs: Text prompts that include your trigger word, plus any style or modifier words.
  • Outputs: Images generated in your custom style or with your trained subject.
    Check the model card for supported resolutions, example prompts, and any quirks or extra parameters.

How can I publish my fine-tuned model?

When your training finishes, you can keep the model private or make it public.
Public models appear in the Use FLUX fine-tunes collection. You can also set pricing and versioning, so others can use your model or build on it.

Can I use these fine-tuned models commercially?

Yes—but follow the license for the base black-forest-labs/flux-1.1-pro model and any terms attached to the fine-tune.
Commercial use is generally allowed when running on Replicate, but some restrictions may apply for local use. Always check the model card for license details.

What should I keep in mind when using FLUX fine-tunes?

  • Include the trigger word in prompts to get the expected results.
  • More and better training images improve fidelity and consistency.
  • Fine-tuned models can behave unpredictably for very unusual prompts.
  • Test a few prompt variations to understand the model’s sweet spots.
  • Review licensing carefully if your fine-tune involves brand assets, real people, or commercial work.