Three few-step LoRAs for Qwen Image 2.1 ship within a week of release
linoy_tsaban · x · 2026-09-25
Less than a week after Qwen Image 2.1's release, three few-step LoRAs have shipped. Viggle's turbo LoRA uses DMD distillation to cut the 40-step base to 6 passes for generation and instruction editing (1-3 reference images), roughly 5x faster with near-parity quality — only dense small text lags (8 steps narrows it). Ships as rank-256 (1.3GB) and rank-128 (680MB) adapters with ComfyUI nodes, ready-made workflows, and a 32-example seed-matched comparison Space. Pruna also released 5- and 8-step variants.
Related event: Viggle's Distilled Qwen-Image-2.1 Generates 5x Faster in 6 Steps(2 posts)→
More from Multimodal
- Midjourney SREF pick: --2523103199 for loose pencil-sketch portraits — tisch_eins · 2026-11-11
- SenseTime open-sources SenseNova 6.8 Code2Video for code-editable video generation — Aiden_Tech_Ai · 2026-11-11
- Gemini-generated endless loop of quantum cryostats self-assembling — hapticdata · 2026-11-11
- Dad brings his daughter's crayon drawings to life with AI for $1.75, and it's unsettling — Illustrious_Ad_4292 · 2026-11-11
- MOSS-Image-SFT vs MiniMax-H3-Image (Pruned): 1:1 image benchmark across 10 art mediums — SkirtSwimming8950 · 2026-11-11
- UniPath (NeurIPS 2026): adaptively coordinating understanding and generation in one model — jindong_wang92 · 2026-11-11