• High fidelity sound effects
  • Ambient sounds
  • Controllable duration
  • 48 kHz
  • Full parameter fine-tuning

MOSS-SoundEffect

An open source sound effect generation model for content creation. Generate ambient sounds, action effects, animal sounds, and short music clips from Chinese or English descriptions, producing up to 30 seconds of audio at 48 kHz. Built on a DiT and Flow Matching architecture, it supports full parameter DiT fine-tuning on your own data.

Author
OpenMOSS Team
Version
v2.0
Open source license
Apache 2.0

MOSS-SoundEffect goes beyond generating high quality audio, bringing systematic improvements to the richness of ambient sound, the range of sound types, and duration control. Trained on more than 10,000 hours of high quality data, it reliably generates natural soundscapes, urban ambience, animal sounds, human actions, and music clips from text prompts. It serves content creation, games, film and television, data synthesis, and other use cases.

The model shares the same architecture as MOSS-TTS.

View architecture

Model Demos