🎨 MiniMax is preparing H3, a new model for video generation.
MiniMax presented the multimodal H3 model for video generation at the WAIC 2026 conference. The model supports the creation of 2K-resolution videos up to 15 seconds long with integrated audio (dialogue, sound effects, and lip-sync). A key feature is the Omni Reference system, which allows for maintaining character consistency.
🌍 The emergence of H3 intensifies competition in the high-quality video model segment (alongside Kling and Veo), offering longer seamless clips and built-in audio processing, which reduces post-production costs.
👤 It is now possible to create longer and more stable videos with the same characters, without spending time on stitching together multiple short fragments and separate voiceovers.
Source 1: https://www.woofun.ai/en/flash/detail/33502 Source 2: https://apidot.ai/blog/minimax-h3-review