Zum Inhalt

Roadmap

VoiceHub 0.3 is in release-hardening mode. The project is not accepting another model family into the built-in registry until the current TTS, ASR, and VAD surface has passed the release-readiness gates.

Current milestone: 0.3 release candidate

The milestone has four ordered outcomes:

  1. Keep source, documentation, benchmark evidence, Git tags, and distribution metadata on one version.
  2. Prove that the wheel and source distribution install in clean environments and contain every registered runtime and required data file.
  3. Separate all-provider contract coverage from the smaller set of public checkpoints that have run end to end on recorded hardware.
  4. Publish only the artifacts that passed the release workflow, using PyPI Trusted Publishing and a protected GitHub pypi environment.

New providers, unmeasured performance claims, and approximate optimizations as defaults are outside this milestone.

Completed scope discovery

The registry currently exposes 68 integrations: 34 TTS, 23 ASR, and 11 VAD. Each has generated documentation, lazy construction, a normalized task output, an explicit training boundary, and a default-installation contract.

The historical GitHub roadmap issue #14 listed LLaSA, Chatterbox, ConversationTTS, CosyVoice, F5-TTS, GPT-SoVITS, Kokoro, MeloTTS, OpenVoice, OuteTTS, Parler-TTS, StyleTTS 2, Orpheus, Dia, and Vui. Those families are now integrated. The issue is retained as historical context; updating or closing it is an external maintainer action, not a release automation step.

After 0.3

Post-release work must start from user demand and reproducible evidence. A new provider needs the same package, license, checkpoint, contract, and hardware evidence as the current registry. A serving or optimization feature needs a measured quality boundary before it can become an automatic default.