NewsBite analysis · AI, Tools · New Product, Software Release
MarkTechPost
An Open-Weights Music Model Generating Complete Five-Minute Songs From Lyrics and a Structured Caption
The recent release of MiniMax-Music3 marks a significant advancement in text-to-music technology, allowing users to generate complete songs up to five minutes long from specified lyrics and structured metadata. This model integrates a Hybrid-LM architecture, enabling efficient music synthesis while supporting commercial usage under specific conditions. MiniMax also provides necessary resources such as weights and inference codes for immediate deployment.
Act nowConfidence: HighImmediate horizon
Published 17 Aug 2026, 18:40 UTC3 min read
Visit original article ↗Lead image — MarkTechPost
Summary
TL;DR
- Generates complete five-minute songs from lyrics and structured captions
- Utilizes a Hybrid-LM combining an 8B Global LLM and a 0.6B Local LLM
- Immediate deployment with provided weights and inference code
- Structured captions enhance the generation process for diverse applications
- Commercial usage granted with visible attribution and limitations for high-revenue users
Interesting facts
- The model merges continuous synthesis techniques with a flow-matching module.
- Synthesis runs on fused continuous hidden states without using a discrete tokenizer decoder.
- MiniMax-Music3 supports a variety of applications including background music and adaptive game soundtracks.