flux.1-kontext-dev
by bfl.ai
Pricing
Input
$0.00 / 1M tokens
Output
$0.01 / 1M tokens
FLUX.1 Kontext [dev] by Black Forest Labs
Overview
FLUX.1 Kontext [dev] is the open-source development variant of the FLUX.1 Kontext multimodal image editing suite from Black Forest Labs, released recently with 12 billion parameters. It supports both text + image input, providing powerful, iterative, context-aware image editing entirely on local machines (huggingface.co).
Features
- Multimodal editing: Text-guided edits applied to existing images
- Iterative editing with strong character/ composition consistency
- Style transfer and selective modification (e.g. lighting, objects, text)
- Fast local inference, optimized for NVIDIA RTX GPUs via TensorRT
- Runnable in ComfyUI and delivers easy-to-use workflow nodes
Use Cases
- Precision image editing—alter hair color, clothing, text, or backgrounds
- Character design—maintain consistency across multiple edits
- Style adaptation—apply watercolor, sketch, cinematic lighting, retaining composition
- Interactive prototyping and iteration using local compute resources
Limitations
- Non-commercial license—open-weight but restricted to non-commercial use
- Model artifacts may appear after repeated refinement — performance lags behind Pro/Max
- Requires ComfyUI nightly build and compatible CUDA/TensorRT setup
Access & Installation
- Available on Hugging Face:
black-forest-labs/FLUX.1-Kontext-dev - Supported in ComfyUI via dedicated workflow nodes
- Optimized inference via TensorRT on NVIDIA RTX GPUs—local performance boost
Quick Comparison
| Feature | Kontext [dev] | Kontext [pro/max] |
|---|---|---|
| Weight access | ✅ Open-source | ❌ Closed / API-only |
| Commercial use | ❌ Non-commercial license | ✅ Commercial licensing available |
| Consistency | High, but with minor artifacts | Superior across multi-step edits |
| Hardware | RTX + TensorRT optimized | API handles backend compute |