Black Forest Labs releases Flux 3 Image: bounding-box + 10 reference images + 4K multi-step local editing
BFL released Flux 3 Image: native bounding-box composition, up to 10 reference images, up to 4K output, and multi-step local editing — pulling controllable generation from concept to workflow.
Source: AI 日报（ai6s.net）
What it is
On October 3, Black Forest Labs (BFL) released Flux 3 Image. The core upgrades over Flux 1.1 sit on controllability and workflow:
- Native bounding-box support: precisely specify subject position and size beyond the prompt
- Up to 10 reference images: single generation can take multiple references at once — unified style, character consistency, viewpoint
- Up to 4K output: direct high-res output, no post-upscale
- Multi-step local editing: repeated local edits on the same canvas without breaking global consistency
Why it matters
Controllable generation previously leaned on external modules (ControlNet, IP-Adapter, etc.) with parameters scattered across models. Flux 3 bakes them into the base — friendlier for production: one prompt path covers composition, reference, resolution and local editing, instead of the "base → inpaint → upscale" three-stage duct tape.
Comparison (image generation)
| Dimension | Flux 3 Image | Qwen-Image-2.1 | Stable Diffusion 3.5 |
|---|---|---|---|
| Native bbox | ✓ | Partial | ✗ (needs ControlNet) |
| Multi-reference | up to 10 | 1–2 (external) | ✗ |
| Output resolution | 4K | 2K | 1K (external upscale) |
| Local editing | Native multi-step | Partial | inpaint as separate pipeline |
Local deployment angle
The Flux series has long supported diffusers / ComfyUI / Forge. 4K output and multi-reference are VRAM-hungry — 24 GB+ consumer or 16 GB+ workstation recommended; quantization variants await community feedback.