🔊 MOSS-SoundEffect v2.0

Text-to-sound-effect diffusion — DiT + Flow Matching, DAC VAE, Qwen3 text encoder. Up to 30 s of 48 kHz audio from a natural-language caption.

Bilingual prompts (English / Chinese) · ZeroGPU

Sound description

Duration (seconds)

The model denoises a fixed 30 s latent and crops to this length.

1 30

Generated audio

Ready.

Example prompts