From the model card
Fields below are copied from the tags and counters on the HuggingFace repository Comfy-Org/flux1-dev at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.
- Publisher (HF namespace)
- Comfy-Org
- License tag
other— read the license file in the repo before relying on it- Lineage
-
- base model black-forest-labs/FLUX.1-Canny-dev
- fine-tune of black-forest-labs/FLUX.1-Canny-dev
- Downloads (HF counter at last fetch)
- 400,167
- Likes (HF counter at last fetch)
- 659
- Model card
- https://huggingface.co/Comfy-Org/flux1-dev
Use cases
- High-fidelity image generation inside ComfyUI workflows
- Generating images with accurate in-image text and typography
- Anatomically consistent character and portrait generation
- Base checkpoint for FLUX LoRA fine-tuning experiments
- Comparing FLUX architecture quality against SDXL checkpoints
Pros
- Rectified flow training produces cleaner gradient paths and sharper outputs than DDPM-based models
- Better text rendering inside generated images than most SD-family models
- Single-file format simplifies ComfyUI installation
- Backed by Black Forest Labs (original Stable Diffusion team)
Cons
- FLUX.1-dev license prohibits commercial use; FLUX.1-schnell or Pro required for commercial projects
- High VRAM requirements (≥16GB recommended for full quality)
- ComfyUI-specific packaging; requires adaptation for Automatic1111 or other UIs
- Slower inference than distilled FLUX variants