Viggle Turbo v0.3 distills Qwen-Image-2.1 to 6 steps, about 5x faster
linoy_tsaban · x · 2026-10-02
Viggle released Turbo v0.3 for Qwen-Image-2.1, a few-step distillation generating in 6 steps instead of 40 (5x faster) with quality nearly indistinguishable from the base model. A new 9-step mode (LoRA off for the last two steps) improves fine detail and small text at 3.5x speedup. v0.3 has less grain than v0.2.1 but slightly softer fine texture; v0.2.1 remains available. Supports ComfyUI via LoRA or single-file int8/fp8/GGUF.
Related event: Viggle Turbo v0.3 Generates Qwen Images in 6 Steps, About 5x Faster(3 posts)→
More from Multimodal
- OneTrainer LoRA recipe: Prodigy optimizer plus a 3-version merge trick — jiggydancer · 2026-10-02
- NoSpoon Studios offers free AI music videos and microdramas on its last public day — Kyrannio · 2026-10-02
- 36 interactive and playable "machines of generative art" coming to a living Generative Art Timeline — Merzmensch · 2026-10-02
- Remocn Studio: free open-source macOS app turns Claude, Codex or Grok into your video editing agent — Scobleizer · 2026-10-02
- Eleven v4 rolls out: new architecture performs scripts with human nuance, free for now — charis_ai · 2026-10-02
- Tavus' Griffin: 48% mistook its AI video avatar for a human on a 1-minute call — The Decoder · 2026-10-02