Token counter
Exact GPT-4o token counts, per-model cost, and how much of each context window a prompt actually fills.
/tokens
Voice studio · runs in your browser · no upload
Free text to speech in your browser
Give your words a voice. Kokoro runs right in your browser: pick a voice, set the pace, and download the audio. No server, no key, nothing uploaded.
TTS voice studio
Say it the way you would say it to a person. Your words are matched to one of 18 tools by readable rules, and the tool runs right here in your browser. No upload, no account.
This tool runs in your browser and needs JavaScript turned on.
esc
How to use the TTS voice studio
- Type or paste your text into the “Text to speak” box. The studio opens with its settings ready, so there is nothing to find first.
- Choose one of 28 American or British voices and a speaking speed from 0.5× to 2×.
- Press “Run /tts”. The first run downloads the voice model once (92 MB); after that, speech starts straight away and works offline.
- Listen in the player and download the WAV file, or save the voice and speed as a named preset for next time.
Things you can type: read it aloud read it aloud at 1.2x
Runs in your browser
The studio runs Kokoro, an open 82-million-parameter voice model, inside this page with kokoro-js. The first time you generate speech, the browser downloads the model from Hugging Face and keeps it in its own storage. Later runs start straight away and work offline.
The standard model is a 92 MB download and runs on the CPU with WebAssembly in every modern browser. The high-quality model is 326 MB and runs on your graphics card with WebGPU, where the browser supports it.
Build a reusable voice
Choose one of 28 English voices with American or British accents, and adjust the speaking speed. Name the recipe and save it in your browser, or export it as JSON for another device. Importing a recipe fills the settings without generating anything.
Voice building here means choosing and saving settings. It does not train a model or clone a recorded speaker.
Listen, compare and download
Generate a sample, listen with the audio player, and download it as a WAV file. Clone a completed run to change the voice or speed while keeping the same text. Each run stays in its own result card for comparison.
Long text is spoken sentence by sentence and joined into one file. Generation time depends on your device; a WebGPU graphics card is usually several times faster than the CPU.
Where your text goes
Nowhere. The text is turned into speech inside this browser tab and is never uploaded. The only download is the model itself, fetched once from huggingface.co.
Saved voice presets live in this browser. Clearing the site’s data removes them, along with the cached model.
Questions
TTS voice studio: common questions
Do I need a server or an API key?
No. The voice model runs in your browser. The first run downloads it (92 MB for the standard model) and later runs reuse the saved copy.
Can I save a custom voice?
Yes. Save a named recipe containing the voice, speed and model. This saves a configuration, not trained voice weights.
Is my text uploaded?
No. Speech is generated in this tab, and saved or exported presets contain voice settings only, never your text.