F5-TTS community API proposal: paired reference audio/text and Chinese MLX path
Proposing a community documentation example for the Chinese F5-TTS MLX integration exposed by Rapid-MLX's local speech API.
The important API distinction is that F5 requires both reference audio and its transcript. The adapter tests cover the paired fields, base64 request encoding, stereo-to-mono handling, and rejection of silent or invalid references. This path uses the f5-tts-mlx implementation; it should retain credit to that port and SWivid's original model work.
The recipe should use an authorized sample with a known transcript and record the generated waveform's sample rate/duration, exact pack and runtime. It should not copy an IndexTTS reference-audio-only request or present the alias as qualification of every F5 language/model release. No synthesis-quality result is claimed in this scope proposal.
Concrete catalog entry: f5-tts-zh → lucasnewman/f5-tts-mlx@2d719cadec8fd3887c8599475a5f65924d523654. This is a conversion/distribution artifact; it is not presented as an official upstream release.
Source references (fixed Rapid commit): catalog, F5 request and waveform contracts, audio guide.
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.