Pika open-sources H3 Flex, a 6-step video model with synchronized audio
huggingface · x · 2026-10-30
Pika Labs has released its first-ever open-weights model, Pika H3 Flex, open source from day one: Apache-2.0 code on GitHub, weights on Hugging Face, and day-one support in ComfyUI and diffusers.
Key details:
- A post-trained MiniMax H3 built for professional creators, focused on control and iteration speed
- Six-step generation, Omni-reference support, native 480p/768p, up to 4K via the Pika API
- Generates video with synchronized audio
- Ships with a quickstart and prompt guide
Related event: Pika releases H3 Flex, its first open-weight video model(19 posts)→
More from Multimodal
- Vogue shoot rigs 70 iPhone 18 Pros for crispy 4D gaussian splatting — bilawalsidhu · 2026-10-30
- A 'Modern Day Horror Story' short made entirely with HeyGen — toolstelegraph · 2026-10-30
- Two Reference Images Fix Generic AI Content in Flows Generations — Aiden_Tech_Ai · 2026-10-30
- SODA CANDY: a wordless 5-minute AI short film one creator spent 16 months making solo — Superb-Meeting7108 · 2026-10-30
- Liquid AI launches LFM3 model family; 1.7B hits 185 tok/s on-device — JosephJacks_ · 2026-10-30
- Open-source ComfyUI film studio turns one prompt into an editable, refinable movie — Goat_bless · 2026-10-30