v0.7.1

What's Changed

• Add Echo-TTS (community model) by @5uck1ess in #180
• Add Parakeet TDT VAD chunking by @0xShug0 in #318
• Accept PCM8/32, float64, A-law, mu-law and WAVEFORMATEXTENSIBLE in the WAV reader by @5uck1ess in #319
• Add Qwen3 ASR word timestamp diagnostics by @0xShug0 in #304
• cuda: zero the MMQ stream-k fixup staging buffer by @iamwavecut in #320
• Improve VibeVoice 7B support by @0xShug0 in #326
• server: resolve the model contract once per model, not per request by @XythQ in #328
• minimax_music3: performance pack — native rope/SwiGLU lowering + opt-in CFG reuse, chunk hop, batched ensemble takes, q4_k depth by @iamwavecut in #321
• Add VoxCPM v1 — lightweight VoxCPM TTS support (0.5B) by @jasonchen31 in #256
• Fix CUDA stream lookup with dynamic ggml backends by @0xShug0 in #340
• docs: update prebuilt binaries section by @drzsdrtfg in #342
• Promote shared MOSS runtime pieces by @0xShug0 in #332
• fail when the cli is given an unknown option by @avionicharshit-byte in #339
• fix: re-feed persistent prefill inputs on cached-graph recompute in QwenCausalDecodeRuntime by @drzsdrtfg in #331
• feat: add Soprano TTS as community model with GGUF packages by @drzsdrtfg in #323
• Honor MiniMax Music3 component options during load by @0xShug0 in #343
• Make VoxCPM2 UI chunking control framework chunks by @0xShug0 in #344
• fixes issue #345 by @huyndao in #346
• Fix Music3 contract spec validation by @0xShug0 in #354
• soprano_tts: port the full reference clean_text normalizer (fixes voiced '.'/dot artifact) by @drzsdrtfg in #355
• feat: Add Audio8-ASR-0.1B community model port (audio8_asr) by @gqf2008 in #337
• Tighten Silero VAD auto-detection by @0xShug0 in #357
• Fix Chatterbox request sequence voice references by @0xShug0 in #379

New Contributors

• @iamwavecut made their first contribution in #320
• @XythQ made their first contribution in #328
• @avionicharshit-byte made their first contribution in #339
• @huyndao made their first contribution in #346

Full Changelog: v0.7.0...v0.7.1