Repository navigation
Ventriloquist 0.1.0
The first release.
- A bench for open voice models that run on Apple silicon: text to speech, voice cloning,
voice conversion and spoken conversation, with every take's speed and memory beside it. - 23 models in the catalog, each downloaded only when you ask for it.
- Built-in voices with previews, cloning from a sample, and direction where a model takes it.
- Conversation with Sophon Realtime 32B Flash, ported to Apple silicon, and tool calling
with LFM2.5 Audio. - Enhance and Scene: a local language model that marks up a line, or writes dialogue from a
direction, for the voice model that will speak it. - Updates itself from this release on.