flux-3-image-klein-comfy
blackForestLabs/flux-3-image-klein-comfy
- Input modalities
- image
- Output modalities
- image
- Capabilities
- imageOutput, visionInput
FLUX models generate and edit images and video, render synchronized audio, and drive robotics. State-of-the-art quality and creative control, through one API.
Source ↗ · 2026-11-12T03:22:35.295203776Zdegraded
BFL generation services and regional APIs
Official status page ↗63 indexed models
63 matching models
blackForestLabs/flux-3-image-klein-comfy
blackForestLabs/flux-3-image-klein-pup
blackForestLabs/flux-tools-deblur-v1
Submits a deblur task. The input image is regenerated by a FLUX blur-removal model to sharpen it while preserving the original scene, objects, composition, and lighting. No prompt needed.
blackForestLabs/flux-tools-erase-v1
Submits an erase task using an input image and a mask identifying the object or region to remove.
blackForestLabs/flux-tools-outpainting-v1
Submits an outpainting task. The input image is placed on a (width, height) canvas at the given offset, and the surrounding region is generated by a FLUX outpainting model.
blackForestLabs/flux-tools-video-edit-v1
Submits a video edit task: the supplied clip is transformed according to the edit instruction. Duration, resolution, aspect ratio, and audio follow the source clip; inference controls are pinned server-side.
blackForestLabs/flux-tools-vto-v1
Submits a virtual try-on task. Person and garment images are mapped to the underlying input image slots; prompts steer attribute transfer.
blackForestLabs/flux-tools-vto-v2
Submits a virtual try-on task against the v2 model. Identical request shape to /vto-v1, with reference and output resolution supported up to 4MP. Person and garment images are mapped to the underlying input image slots. An edit instruction is generated from the person and garment
blackForestLabs/flux-tools-video-upscale-v1
Submits a video upscaling task: 1.5x-3x super-resolution of the source clip (up to 2560x1440 in, 13.75 MP output frames). The upscale covers the first 20 seconds of the source; clips well past that are rejected. `creativity` selects precise source-faithful upscaling (0) or creati
These charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.