OpenAI-compatible TTS endpoint using OmniVoice: 0.3s response time
A new OpenAI-compatible TTS server, utilizing OmniVoice, offers rapid speech generation with a response time of approximately 0.3 seconds on an RTX 3080. This server can clone voices from short reference clips and is optimized for quick generation, processing long texts paragraph by paragraph. The developer uses it to convert school books into audiobooks in their own voice, enabling listening while driving. More details are available on the project page.
This new OpenAI-compatible TTS server achieves a 0.3-second response time, a speed benchmark for voice cloning and text-to-speech generation unlike many other solutions.
Time & source
Times shown in UTC
Display time zone: UTC
Local time zone unavailable; showing UTC.
PublishedOffset at this time: UTC+0Oct 8, 2026, 16:31 UTC
IngestedOffset at this time: UTC+0Oct 9, 2026, 02:00 UTC
- Published
- Oct 8, 2026, 16:31
- Ingested
- Oct 9, 2026, 02:00
- Source type
- Dev community
- Tier
- Community
- Source status
- Healthy
Tier is a per-source editorial setting, not a per-item score.
Discussion trend
The percentage is based on collected discussion signal, not new comments or independent people. The curve only compares the same topic across time.
I want to share a TTS server with an OpenAI-compatible API that generates speech really fast (about 0.3 seconds for a sentence on an RTX 3080) and can clone a voice from a short reference clip. I’ve optimized the server so generation starts quickly, and it processes long text in sequence, paragraph by paragraph. I use it to turn school books into audiobooks in my own voice, so I can listen to them while driving.
Out of the box, it’s already tuned for the best settings, but you can change them however you like, for example, the CFG (guidance) scale.
Here are the links to the repo and to a page that showcases it, where you can listen to all the voices. As always, it’s open source and free for anyone to use and modify.