AI BriefingPika releases H3 Flex, a post-trained MiniMax H3 model with open weights and 7-second generation speed
·2026.10.29 09:00
Key point
The model supports up to 12 reference files, exports five separate audio stems, and ranks #1 among open-weights models and #7 overall on DesignArena’s text-to-video leaderboard.
Details
- Pika introduces H3 Flex, a post-trained MiniMax H3 model designed for professional iteration, featuring faster generation and open weights.
- The model generates a 15-second clip with synchronized audio at 768p in as little as seven seconds of transformer computation, using six denoising steps instead of the original 50.
- Flex accepts up to 12 reference files (images, video, audio) to guide character, world, and motion, and allows reference-guided editing to modify specific elements while preserving performance.
- Open weights for generator adapters and inference code are released on Hugging Face and GitHub, covering text-to-video, first/last-frame generation, and multimodal reference-to-video at 480p and 768p.
- Pika-hosted features include upscaling to 1080p, 2K, or 4K, and separation of five audio stems (Dialogue, Score, SFX, Foley, Ambience) via the Pika API.
- The model is available via the Pika API Club and will be added to the Pika creation product on November 2, 2026.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
SGLang Achieves Up to 6.24x Speedup in MiniMax-H3 Video Generation Using Cache-DiT and SubBlockFish Audio Drama 3 Ranks #1 Open-Weight Model on Artificial Analysis Speech Arenas