MOSS-TTS-Nano community integration proposal: continuation mode and tokenizer provenance
Proposing a narrowly scoped community note for MOSS-TTS-Nano through Rapid-MLX's local /v1/audio/speech endpoint, using mlx-audio.
An important boundary surfaced in the adapter review: Rapid currently constructs generation kwargs with mode="continuation" and a separately resolved audio_tokenizer_source. That is not the voice-clone workflow emphasized by your official CLI examples. The catalog exposes the default voice and pins the MOSS-Audio-Tokenizer-Nano dependency separately.
A useful integration note should explain that current continuation-only surface, verify a valid waveform and codec dependency, and leave reference-voice support out until it is implemented and exercised. The linked tests document the codec resolution and exact kwargs contract; they do not establish synthesis quality. Would that explicit subset be appropriate as a community API example?
Concrete catalog entry: moss-tts-nano β mlx-community/MOSS-TTS-Nano-100M@229a9c51bb0ffff6fd0dbe53b5bf0c441e438a79. This is a conversion/distribution artifact; it is not presented as an official upstream release.
Source references (fixed Rapid commit): catalog, TTS continuation path, codec and kwargs tests.
Verification status: this proposal is based on reading the catalog, implementation and the linked documentation/tests. No fresh model inference or test-suite run was performed for this submission; source tests are not an end-to-end qualification. It requests a community documentation scope, not an official endorsement or a performance listing.
Disclosure: I maintain Rapid-MLX. Prepared with AI assistance under my direction. Anonymous usage-metadata telemetry is enabled by default; DO_NOT_TRACK=1 disables it. The original model and conversion terms continue to apply.