← Back to Stretch goals

Prepare your guide’s audio

Generate Kokoro recordings here, then use them in any routine. The voice model downloads once and runs in your browser. Keep this page open while preparing audio. Allow roughly 100 MB for the model, plus space for the recordings.

Choose a voice before generating. Existing recordings keep their original voice; clear them first if you want a different voice throughout.

Ready. Nothing downloads until you press a generate button.

Generated clips stay in this browser. Playback uses a matching hosted recording first, then a browser recording, then live speech. Export recordings makes an audio pack for publishing; your routine backup remains separate. Custom instructions are processed locally by Kokoro. The library and model are downloaded from jsDelivr and Hugging Face.