The hub for audio AI research: papers, open models, benchmarks & datasets across audio LLMs, speech recognition, TTS, music & audio generation.
-
Updated
Nov 16, 2026 - Python
The hub for audio AI research: papers, open models, benchmarks & datasets across audio LLMs, speech recognition, TTS, music & audio generation.
An Open-Source Project to Unify Audio Processing and Generation
A speech-to-text system you can train for $25. A frozen speech encoder and a Qwen LLM joined by a small trained projector, with word timestamps, speaker diarization, a batched HTTP server, a live demo, and a free course.
Flyto MLX — Apple Silicon LLM server with audio chat / DFlash / Chinese model presets. Fork of @jundot/oMLX.
免训练方言关键词检索与白盒审计台:80ms/40ms 双网格防泄漏、Top-20 语境重排、Prefix KV-Cache 加速与五折阴性实验公开。Training-free Audio-LLM keyword spotting with a live white-box Audit Studio.
Code, per-clip scores and audit for 'The Representation Utility Gap in Audio LLMs for Clinical Speech' (ICASSP 2027 submission)
Audio-model training support in ms-swift (used for Oylan2)
Bilingual (Arabic/English) multi-speaker conversation analysis with audio language models: Voxtral, pyannote, librosa, Qwen3.
acoustic-abstention: learning when to stay silent, acoustic abstention encoding for ASR. The audio encoder learns to return empty transcripts for non-speech while the decoder stays frozen: hallucination at most 0.34% on UrbanSound8K, WHAM! and FSD50K for Whisper large-v3 and three audio-LLM ASR models, speech WER/CER within 0.5 points
To associate your repository with the audio-llm topic, visit your repo's landing page and select "manage topics."