Two voice engines, one goal: sound human
Kokoro powers the English voices. Kokoro-82M is a compact, openly licensed text-to-speech model that ranks among the most natural-sounding free voices available. There are 20 American and 8 British voices, from the warm Heart and Emma to the deep Fenrir and the storytelling Fable.
Piper covers 48 more languages with over 150 neural voices, including Spanish (Spain, Mexico, Argentina), Portuguese (Brazil, Portugal), French, German, Italian, Hindi, Arabic, Persian, Urdu, Russian, Ukrainian, Polish, Turkish, Chinese, Korean, Vietnamese, Indonesian, Swahili, Bengali, Malayalam, Telugu and many more.
How the voice drives the video
Narration is not laid on top at the end — it sets the timing. Each scene starts with a short breath, plays its voice clip, then holds for a beat before the transition. Captions highlight each word as it is spoken, background music dips automatically under the voice, and when you edit a sentence only that scene is regenerated.
Getting the best performance from an AI voice
- Write how people talk. Short sentences, contractions, everyday words.
- Use punctuation for pacing. Commas create small pauses; full stops create longer ones.
- Spell out tricky terms the way they should sound if the voice mispronounces them.
- Adjust speed between 0.7× and 1.4×. Slightly slower (0.9×) suits lessons; slightly faster suits ads.
- Audition voices with the Hear this voice button, which reads your own first sentence.
Why it is free and private
Cloud voice services charge per character because they run on rented GPUs. These models run inside your browser using WebAssembly (and WebGPU where available). The model downloads once — about 92 MB for Kokoro and 20–75 MB for a Piper voice — and is cached for next time. Your script never leaves your device.
Or use your own voice
Switch to My voice to record each scene with your microphone or upload audio files. This works in every language and accent, and scenes stretch to fit your delivery. It is the best choice for personal stories, languages without an AI voice, and anyone who simply prefers to be heard. Try it in the animated video maker, or browse every voice by language.