Viggle distills Qwen-Image-2.1 to 6 steps, ~5x faster with near-parity quality
RisingSayak · x · 2026-11-10
Viggle released Qwen-Image-2.1-viggle-turbo (v0.3), a few-step distilled LoRA for Alibaba's Qwen-Image-2.1, available on Hugging Face with Diffusers support.
- Text-to-image plus instruction-based editing with 1–3 reference images in 6 steps instead of 40, no classifier-free guidance — roughly 5x faster end to end (4.7s vs 26.1s in the comparison image).
- Quality is highly competitive with the base model; clearest gaps are small dense text and complex edits.
- New 9-step mode: 7 turbo steps, then the LoRA switches off and the base model finishes — finer detail, more accurate small text, still 3.5x faster.
- The team says 6 steps is near the distillation's capacity: sharper comes with grain, less grain with softer look; further gains must be bought with steps.
More from Multimodal
- Is ControlNet Obsolete? New DiT Image Models Ship Without It — Manicarus · 2026-11-10
- Full prompt revealed: how to recreate the Van Gogh convenience store in MiniMax H3 Image — mhdfaran · 2026-11-10
- 10 open-source GitHub repos for voice AI: from local ASR to zero-shot voice cloning — Shruti_0810 · 2026-11-10
- Fable 5.2 generates animated nature scene with ~70k trees from one prompt — Merzmensch · 2026-11-10
- MOSS-Image roundup: Control LoRAs, 461 new characters, and 20-minute LoRA training on a 5090 — optimisticalish · 2026-11-10
- StreamEval from Tsinghua, NVIDIA, CMU and Berkeley exposes gaps in streaming video generation across 38 models — jiqizhixin · 2026-11-10