Adobe's FlowTool frames image retouching as flow matching, cutting latency 50x
adobe-research · hf · 2026-09-29
Adobe Research's FlowTool reformulates tool-based image retouching as conditional flow matching: a VLM backbone plus a DiT parameter generator turns noise into editing parameters, trained with a two-stage supervised curriculum and reward-based post-training. It beats specialized MLLM agents on reference-based metrics across MMArt-Bench, ArtEdit-Bench and MIT-Adobe5K, while cutting latency by at least 50x and memory by 2x.
More from Multimodal
- Seedance 2.5 single-prompt video turns the Shire into a gritty selfie vlog — techhalla · 2026-09-29
- Reddit user shows zero-cut H3 long video with dynamic lighting, dares you to find the seam — SIR_NVAX_A_LOT · 2026-09-29
- Match cuts with generative AI: transform elements mid-action while keeping motion continuity — ArcaArtificial · 2026-09-29
- invideo launches agentic video editor that edits a real multitrack timeline — azed_ai · 2026-09-29
- Redditor Iterates a Full Show Episode with Opus 5.5 — No One-Shot, but Pretty Happy — weakcper · 2026-09-29
- MiniMax H3 used to create animated Lord of the Rings demo — azed_ai · 2026-09-29