Kokoro TTS - Voice Comparison
Free, open weight models can generate speech. On your local machine.
Listen to voice samples of all Kokoro TTS voices on the web.
Optionally, install Kokoro with Podman.
Listen and Compare Voices in Your Browser # ↑
Search and filter by gender and language. The demo text is currently only in English.
Kokoro is a free text-to-speech engine, architected by Li, Han, Raghavan, Mischler and Mesgarani; and trained by @rzvnzn. I just created the samples and the page to filter them. Their default flagship voice is af_heart (English, female), but there are many other interesting voices.
Run Kokoro with Podman # ↑
These commands assume you already have Podman installed. Podman runs containers like Docker. It's safer than Docker, as it does not require root (or the infamous root-equivalent docker group).
Let's download and run Brauns ready-made Kokoro container:
$
Web user interface: http://localhost:8880/web/
API, usable with Python requests http://localhost:8880/v1
See also # ↑
Hacker News 2026: Local, CPU-Friendly, High-Quality TTS (Text-to-Speech) with Kokoro
Li, Han, Raghavan, Mischler and Mesgarani, @rzvnzn 2026: hexgrad/Kokoro-82M on HuggingFace
Braun 2026: remsky / Kokoro-FastAPI on Github
