Unified video-audio generation eliminates lip-sync rework. Production cycle visibly faster.
MiniMax H3 Now Available in Luma Agents
Original: MiniMax H3 Available Now | Luma
Importance: 新規ビデオ生成モデルの段階的GA で、ユーザーサブセットに影響。ワークフロー簡素化は実装価値があるが、エンタープライズ利用には時間差あり。
Summary
Luma Labs launched MiniMax H3 in Luma Agents, enabling unified video and audio generation from text, images, video, and audio references in a single pass. Previously requiring separate steps (generate video, find voice, mix), H3 eliminates handoffs and lip-sync drift by producing dialogue, sound effects, and ambience pre-timed to on-screen action. Available now for self-serve accounts as of August 5.
Key Points
- Single-pass generation from text, image, video, and audio inputs to synchronized video + audio output
- Consolidates three-step workflow (video → voice selection → mixing) into one inference
- Eliminates lip-sync drift and re-work cycles, accelerating production speed
- Rolled out to self-serve accounts first; enterprise access planned incrementally
View developer notes (APIs, breaking changes, migration)
MiniMax H3 available via Luma Agents platform, accepting multi-modal inputs (text prompt, keyframe image, reference video, voice memo) and returning video + audio + dialogue + SFX + ambience in single inference pass. Architecture merges video generation, speech synthesis, and orchestration layers, eliminating separate voice-selection and mixing steps. Self-serve tier only (as of Aug 5); enterprise rollout planned. Single-pass generation reduces lip-sync drift from inter-step handoffs.
Source: https://lumalabs.ai/news/minimax-h3-now-available-in-luma-agents
Outlet: Luma Labs
This article is an AI-generated summary (OpenAI GPT-4o-mini) of publicly available information from Anthropic, OpenAI, Google, Meta, Mistral, DeepSeek, Sakana, and other vendors. The original source URL is always provided in accordance with fair-use citation requirements. Summaries are AI-generated and may contain mistranslations or misinterpretations. Always verify details with the original source.