Quanta AI Labs

Xiaomi · Audio Processing

MiDashengLM-Gen

Generates whole audio scenes - speech, music, effects and ambience mixed together - from a text description at 16kHz, rather than treating speech synthesis as a separate problem. Runs a fully fine-tuned Qwen3-1.7B backbone over a DashengTokenizer producing 768-dimensional latents at 25Hz, with a 16-layer Flow Matching DiT and a learned stop head. Apache-2.0.

Open the source ↗

Similar Audio & Speech releases

  • NVIDIA Nemotron 3.5 Transcribe 30B-A4B

    NVIDIA · Open multilingual ASR · Week 45, 2026

  • Who

    Desert Ant Labs · On-device diarization · Week 45, 2026

  • Dessa

    Block (Dessa research lab) · Speech-only conversation model · Week 45, 2026

  • WorldSonus

    NoizAI · Real-Time Video-to-Audio · Week 42, 2026

  • Phonon-2

    Fermion Research · Compressed English ASR · Week 41, 2026

  • Gemini 3.8 Live

    Google · Live Voice Agent · Week 39, 2026

More from Xiaomi

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.