Repository Issue Activity (beta)

#903 Model request + implementation plan: dots.tts (rednote-hilab) zero-shot voice cloning

#902 STT load_model silently returns a randomly initialized model when checkpoint keys don't match (strict=False default)

#900 Voxtral: 32000 is used as an EOS token but it is the text token " Capital", silently truncating transcripts

#898 Streaming `/v1/audio/speech` emits one complete container file per chunk, inserting audible silence at every seam (all container formats)

#892 Qwen3-TTS-12Hz-1.7B-Base-bf16 voice selector not works

#882 Qwen3-TTS audio gets early silent EOS truncation at sentence tail even with max_tokens=8192

#890 Dia: codebook_to_audio discards ~175ms of real audio from the end of every generation

#854 convert: non-deterministic model type detection — exact scoring tie makes Higgs Audio v3 conversion randomly fail

#874 [Qwen3-TTS] float32 speaker embedding silently promotes the whole talker prefill + KV cache to float32, ~2.4x slower decode on bf16 checkpoints

#875 POST /v1/audio/transcriptions: ndjson streaming reports inference-time failures as 200 with an empty body

#870 STT downsampling has insufficient alias rejection since #825

#846 moss_tts_delay: shape mismatch loading mlx-community/MOSS-TTS-Local-Transformer-MLX-8bit

#847 Add MLX port for Chatterbox Multilingual v3

#849 transformers version compatibility

#871 Model Request: nvidia/NVIDIA-NemotronLabs-VoiceChat-11B

#753 Feature: Add SenseVoice STT backend with emotion and audio-event detection

#736 VoxCPM2-8bit still not loadable in 0.4.3; add voxcpm2 loader support

#851 Qwen3-TTS: mx.clear_cache() in streaming path costs ~13% RTF on M1 Pro

#339 Feature Request: Hotwords support for FunASR model

#44 how to use sesame labs?

#16 Support Stable Audio Open v1.0

#280 Voices in UI and available models

#28 Training?

#413 Got error "fix_mistral_regex=True " on Qwen3-TTS-12Hz-1.7B-VoiceDesign-bf16

#101 Crash when running Dia-1.6B