black-forest-labs/flux-dev-lora 🔢🖼️📝✓❓ → 🖼️

⭐ Official ▶️ 4.8M runs 📅 Nov 2024 ⚙️ Cog 0.15.8 🔗 GitHub ⚖️ License

image-to-image lora lora-inference text-to-image

About

A version of flux-dev, a text to image model, that supports fast fine-tuned lora inference

Example Output

Prompt:

"style of 80s cyberpunk, a portrait photo"

Output

Performance Metrics

2.83s Prediction Time

2.84s Total Time

1 Images

All Input Parameters

{
  "prompt": "style of 80s cyberpunk, a portrait photo",
  "go_fast": true,
  "guidance": 3,
  "lora_scale": 1,
  "megapixels": "1",
  "num_outputs": 1,
  "aspect_ratio": "1:1",
  "lora_weights": "fofr/flux-80s-cyberpunk",
  "output_format": "webp",
  "output_quality": 80,
  "prompt_strength": 0.8,
  "num_inference_steps": 28
}

Input Parameters

seed Type: integer: Random seed. Set for reproducible generation
image Type: string: Input image for image to image mode. The aspect ratio of your output will match this image
prompt (required) Type: string: Prompt for generated image
go_fast Type: booleanDefault: true: Run faster predictions with model optimized for speed (currently fp8 quantized); disable to run in original bf16. Note that outputs will not be deterministic when this is enabled, even if you set a seed.
guidance Type: numberDefault: 3Range: 0 - 10: Guidance for generated image
extra_lora Type: string: Load LoRA weights. Supports Replicate models in the format <owner>/<username> or <owner>/<username>/<version>, HuggingFace URLs in the format huggingface.co/<owner>/<model-name>, CivitAI URLs in the format civitai.com/models/<id>[/<model-name>], or arbitrary .safetensors URLs from the Internet. For example, 'fofr/flux-pixar-cars'
lora_scale Type: numberDefault: 1Range: -1 - 3: Determines how strongly the main LoRA should be applied. Sane results between 0 and 1 for base inference. For go_fast we apply a 1.5x multiplier to this value; we've generally seen good performance when scaling the base value by that amount. You may still need to experiment to find the best value for your particular lora.
megapixels Default: 1: Approximate number of megapixels for generated image
num_outputs Type: integerDefault: 1Range: 1 - 4: Number of outputs to generate
aspect_ratio Default: 1:1: Aspect ratio for the generated image
hf_api_token Type: string: HuggingFace API token. If you're using a hf lora that needs authentication, you'll need to provide an API token.
lora_weights Type: string: Load LoRA weights. Supports Replicate models in the format <owner>/<username> or <owner>/<username>/<version>, HuggingFace URLs in the format huggingface.co/<owner>/<model-name>[/<lora-weights-file.safetensors>], CivitAI URLs in the format civitai.com/models/<id>[/<model-name>], or arbitrary .safetensors URLs from the Internet, including signed URLs. For example, 'fofr/flux-pixar-cars'. Civit AI and HuggingFace LoRAs may require an API token to access, which you can provide in the `civitai_api_token` and `hf_api_token` inputs respectively.
output_format Default: webp: Format of the output images
output_quality Type: integerDefault: 80Range: 0 - 100: Quality when saving the output images, from 0 to 100. 100 is best quality, 0 is lowest quality. Not relevant for .png outputs
prompt_strength Type: numberDefault: 0.8Range: 0 - 1: Prompt strength when using img2img. 1.0 corresponds to full destruction of information in image
extra_lora_scale Type: numberDefault: 1Range: -1 - 3: Determines how strongly the extra LoRA should be applied. Sane results between 0 and 1 for base inference. For go_fast we apply a 1.5x multiplier to this value; we've generally seen good performance when scaling the base value by that amount. You may still need to experiment to find the best value for your particular lora.
civitai_api_token Type: string: Civitai API token. If you're using a civitai lora that needs authentication, you'll need to provide an API token.
num_inference_steps Type: integerDefault: 28Range: 1 - 50: Number of denoising steps. Recommended range is 28-50, and lower number of steps produce lower quality outputs, faster.
disable_safety_checker Type: booleanDefault: false: Disable safety checker for generated images.

Output Schema

Output

Type: array • Items Type: string • Items Format: uri

Example Execution Logs

Lora fofr/flux-80s-cyberpunk already loaded
running quantized prediction
Using seed: 1827486286
  0%|          | 0/28 [00:00<?, ?it/s]
  7%|▋         | 2/28 [00:00<00:01, 18.09it/s]
 14%|█▍        | 4/28 [00:00<00:01, 13.05it/s]
 21%|██▏       | 6/28 [00:00<00:01, 11.97it/s]
 29%|██▊       | 8/28 [00:00<00:01, 11.50it/s]
 36%|███▌      | 10/28 [00:00<00:01, 11.25it/s]
 43%|████▎     | 12/28 [00:01<00:01, 10.86it/s]
 50%|█████     | 14/28 [00:01<00:01, 10.83it/s]
 57%|█████▋    | 16/28 [00:01<00:01, 10.83it/s]
 64%|██████▍   | 18/28 [00:01<00:00, 10.84it/s]
 71%|███████▏  | 20/28 [00:01<00:00, 10.81it/s]
 79%|███████▊  | 22/28 [00:01<00:00, 10.69it/s]
 86%|████████▌ | 24/28 [00:02<00:00, 10.68it/s]
 93%|█████████▎| 26/28 [00:02<00:00, 10.71it/s]
100%|██████████| 28/28 [00:02<00:00, 10.73it/s]
100%|██████████| 28/28 [00:02<00:00, 11.05it/s]
Total safe images: 1 out of 1

Version Details

Version ID: ae0d7d645446924cf1871e3ca8796e8318f72465d2b5af9323a835df93bf0917
Version Created: June 25, 2025

Run on Replicate →