Simplified Matching Distillation (SMD) is introduced for few-step diffusion, matching teacher flow with a straight-line prediction plus a curved zero-integral path at random sub-intervals. It requires no auxiliary score model, no adversarial training, and no generator modification, and claims photorealistic appearance, high-quality dynamics, memory-efficient training, and mode covering.
Key points
No auxiliary score model.
No adversarial training.
Sources
Soundverse Android launch – live with 1-minute 14-second prompt output
Soundverse for Android is live, generating instrumental, title and cover art from a single prompt in 1m 14 s on the phone. The AI studio also offers text-to-music, stem separation, DNA-based scoring and lyric video export, and starts free with 1,000 tokens.
Key points
Android app release: live on device.
One prompt creates instrumental, title and cover art in 1m 14 s.
Free starter allocation of 1,000 tokens.
Features include AI music generator, agent workflows, stem separation, DNA scoring and lyric video export.
Sources
- Original post
- AI Music, Video & Voice Generator - Linked page
