Hugging Face New from Stability AI: Creative Sound Generator
7 snips
Jun 1, 2025 Discover how Stability AI's new mobile music generation model is transforming the creative landscape for musicians. Explore its innovative use of copyright-free resources and how it stacks up against competitors. Learn about the latest advancements, including AI-generated audio for video, and hear insights into the company's challenging yet inspiring journey. The discussion reveals the evolving potential of generative AI in music and multimedia, igniting a conversation on future possibilities.
AI Snips
Chapters
Transcript
Episode notes
Stability AI's Lightweight Music Model
- Stability AI's new audio model generates music, focused on short samples and sound effects rather than vocals.
- It is lightweight, approx 341 million parameters, and can run directly on smartphones with ARM CPUs.
Copyright-safe Music Model Tradeoffs
- Stability AI trained their model only on copyright-clear royalty-free audio to avoid IP issues.
- This makes it less capable compared to competitors that use broader, potentially copyright-containing data.
Model Limitations and Audio Scope
- The model creates up to 11-second audio clips like drums and riffs within about 8 seconds on a smartphone.
- It does not support vocals, has limited styles, and mainly covers Western music due to its training data.