Kandinsky 6.0 Lite (3B) weights land on Hugging Face with Diffusers pipeline
lmoroney · x · 2026-10-13
The Diffusers weights for Kandinsky 6.0 Video Lite (3B) are live on Hugging Face under MIT, with a full model card and runnable code.
- Supports T2AV / TI2AV pipelines (Kandinsky6TI2VAPipeline): 5-second clips with synchronized 44 kHz audio including lip-sync
- Example usage: bfloat16 loading + enablemodelcpuoffload, feasible on consumer hardware
- Prompts can choreograph audio cues (ambience, SFX, score with timing) inline alongside the visual description; a super-resolution add-on scales output to 1080p
- Paper: arXiv 2610.05608
The sample prompt — visual description plus an "Audio: ..." block in one string — is the key pattern for getting started.
Related event: Kandinsky 6.0 Open-Sourced with Synchronized Audio-to-Video Generation(2 posts)→
More from Multimodal
- A realistic-style Midjourney sref code (--sref 2170921320) making the rounds — umesh_ai · 2026-10-13
- Prompting Opus 5.5 with Blender to build a ramen shop yields "unmatched" 3D results — victormustar · 2026-10-13
- Voyager launches as an open harness driving Opus, Astra and DeepSeek across Blender, Resolve and 100+ creative tools — ravisparikh · 2026-10-13
- toyxyz open-sources dual Anima ControlNet-LLLite adapters for 3D pose and region control — bdsqlsz · 2026-10-13
- Midjourney sref 3978901914 yields clean fine-pen silhouette illustrations — michaelrabone · 2026-10-13
- Horror-cabaret AI music video made with Claude, Grok Imagine, Midjourney and Suno — Match_Living · 2026-10-13