prunaai/flux-kontext-fast 🔢📝❓🖼️ → 🖼️

⭐ Official ▶️ 17.3M runs 📅 Jun 2025 ⚙️ Cog 0.16.0 🔗 GitHub 📄 Paper ⚖️ License
image-editing image-to-image

About

Ultra fast flux kontext endpoint

Example Output

Prompt:

"Replace “schnell” by “Kontext” and make the prune blue"

Output

Example output

Performance Metrics

1.74s Prediction Time
1.80s Total Time
All Input Parameters 
{
  "seed": -1,
  "prompt": "Replace “schnell” by “Kontext” and make the prune blue",
  "guidance": 2.5,
  "image_size": 1024,
  "speed_mode": "Real Time",
  "aspect_ratio": "match_input_image",
  "img_cond_path": "https://replicate.delivery/pbxt/NFoiYQ8DdIEQFepICB7SpZB6mI2HC4xZjiHl9mXFqMhZS9sY/flux_schnell.png",
  "output_format": "jpg",
  "output_quality": 80,
  "num_inference_steps": 30
}
Input Parameters
seed Type: integerDefault: -1
Seed
prompt (required) Type: string
Prompt
guidance Type: numberDefault: 3.5
Guidance scale
image_size Type: integerDefault: 1024
Base image size (longest side)
speed_mode Default: Extra Juiced 🔥 (more speed)
Speed optimization level
aspect_ratio Default: match_input_image
Aspect ratio of the output image
img_cond_path (required) Type: string
Input image for image to image mode. The aspect ratio of your output will match this image
output_format Default: jpg
Output format
output_quality Type: integerDefault: 80Range: 1 - 100
Output quality (for jpg and webp)
num_inference_steps Type: integerDefault: 30
Number of inference steps
Output Schema

Output

Type: string • Format: uri

Example Execution Logs 
Running prediction with args: ['prompt', 'image', 'height', 'width', 'guidance_scale', 'num_inference_steps', 'generator']
Generation `height` and `width` have been adjusted to 1024 and 1024 to fit the model requirements.
  0%|          | 0/30 [00:00<?, ?it/s]
 50%|█████     | 15/30 [00:00<00:00, 20.99it/s]
100%|██████████| 30/30 [00:01<00:00, 24.64it/s]
100%|██████████| 30/30 [00:01<00:00, 24.01it/s]
Inference took 1.63 seconds
Version Details
Version ID
6efb57153457f8c51fb813c6d15f45d896f1916dd7d732af49d6a4b09488e2a6
Version Created
November 6, 2025
Run on Replicate →