Write a conversation, give each speaker a voice, and get back a natural-sounding podcast you can download — without recording anything or uploading your script. The voices are generated by a compact AI speech model (Kokoro) that runs right inside your browser, so your words never leave your device.
It is built for back-and-forth dialogue: label each line with a speaker name and the tool stitches every line together into one continuous track, with natural little pauses between turns. Great for faceless podcasts, explainers, dialogue practice, and rough voiceovers.
No. The speech model runs entirely in your browser; your script and the audio it produces never leave your device. The model files download once (from this site or a public source) and are cached.
They use Kokoro, a modern on-device model that sounds natural and conversational — a big step past robotic browser voices. Because everything runs locally it is not quite studio/cloud quality, but it is very usable for podcasts, drafts and faceless content.
As many as you like — label each line with a name (Alex:, Sam:, Jordan:) and assign each a voice. American and British male and female voices are available.
The AI model (~80 MB) downloads once and is cached by your browser. Generation is fastest on devices with WebGPU (recent Chrome/Edge); elsewhere it still works, just slower.
Download the WAV to use anywhere — publish it, edit it, or drop it into another tool to add captions or a talking-character video.