This is a simplified guide to an AI model called Flux-Kontext-Fast maintained by Prunaai. If you like these kinds of analysis, you should join AImodels.fyi or follow us on Twitter.

Model overview

flux-kontext-fast is an optimized endpoint for FLUX.1 Kontext, developed by prunaai using the Pruna framework for enhanced speed. This model builds on Black Forest Labs’ image editing capabilities, offering a faster alternative to the standard flux-kontext-dev while maintaining quality. Similar to how flux-schnell optimizes the faster schnell variant and flux.1-juiced optimizes the dev model, this endpoint delivers accelerated performance through intelligent model optimization.

Model inputs and outputs

The model accepts an image along with a text prompt describing desired edits, and returns a modified image. Key parameters control the generation process, including guidance strength, inference steps, and output quality, allowing fine-tuning between speed and result fidelity.

Inputs

• prompt: Text description of the edits or changes to apply to the image
• img_cond_path: Input image for editing mode
• guidance: Guidance scale controlling how closely the model follows the prompt (default: 3.5)
• num_inference_steps: Number of inference steps affecting quality and speed (default: 30)
• seed: Random seed for reproducibility (default: -1 for random)
• image_size: Base image size in pixels for the longest side (default: 1024)
• aspect_ratio: Output aspect ratio, either matching the input image or specific dimensions like 16:9 or 1:1
• speed_mode: Optimization level from “Lightly Juiced” (more consistent) to “Real Time” (maximum speed)
• output_format: File format for the result (png, jpg, or webp)
• output_quality: Quality level for jpg and webp formats (1-100, default: 80)

Outputs

• Output: URI pointing to the generated edited image

Capabilities

This model performs in-context image e...

Click here to read the full guide to Flux-Kontext-Fast