Any native timestamp?

[YacratesWyh]

Hi, as it's generated voice, I think there should be some timestamp that I could use to generate viseme, any suggestions of how to start?

[dosubot]

Hi @YacratesWyh! I'm Dosu and I’m helping the OpenVoice team.

OpenVoice doesn't natively provide timestamps or phoneme alignment info in its generated audio—there's no built-in way to get timing data for viseme sync from the TTS or voice conversion APIs. To generate viseme timing, you'll need to use external forced alignment or phoneme recognition tools. The typical workflow is: generate audio with OpenVoice, then run a forced aligner (like Montreal Forced Aligner, Gentle, or Whisper with timestamps) on the output audio and its transcript to extract phoneme/word timestamps, and finally map those to visemes for animation or lip sync. This is the standard approach since OpenVoice outputs only raw audio without timing metadata.
Reference: OpenVoice API code and issues

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}

   

[YacratesWyh]

MFA has a relative low efficiency, while whisper is not accurate for viseme.