> ## Documentation Index
> Fetch the complete documentation index at: https://docs.origami.techmitten.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Voice Studio: Browser-Based TTS with Kokoro.js Voices

> Generate high-quality text-to-speech audio with dozens of voices and download as MP3 — powered by Kokoro.js, fully in-browser, nothing ever uploaded.

Voice Studio turns any text into natural-sounding speech, entirely inside your browser. Powered by Kokoro.js, every word is synthesized on your device — no audio is sent to any server. Choose from a large library of voices spanning American, British, and European accents, dial in speed and quality, and download a finished MP3 in seconds.

## Generating Audio

<Steps>
  <Step title="Enter your text">
    Type or paste the text you want to convert into speech in the main input area. There is no hard character limit, but shorter passages generate faster and are easier to preview.
  </Step>

  <Step title="Select a voice">
    Pick a voice from the dropdown. Each voice has a distinct accent, pitch, and style — see [Available Voices](#available-voices) below for the full naming guide.
  </Step>

  <Step title="Set speed and bitrate">
    Adjust the **Speed** slider to control how fast or slow the speech is delivered. Choose your target **MP3 bitrate** (128, 192, or 320 kbps) based on the quality you need.
  </Step>

  <Step title="Click Generate">
    Hit **Generate** and Voice Studio synthesizes the audio in your browser. A waveform preview appears when it's ready.
  </Step>

  <Step title="Download your file">
    Click **Download** to save the MP3 to your device. The filename reflects your voice and speed selection for easy organization.
  </Step>
</Steps>

## Available Voices

Voice Studio ships with a large library of voices. The naming convention tells you the speaker's region and gender at a glance:

| Prefix | Meaning         | Example voices           |
| ------ | --------------- | ------------------------ |
| `af_`  | American Female | `af_heart`, `af_bella`   |
| `am_`  | American Male   | `am_adam`, `am_michael`  |
| `bf_`  | British Female  | `bf_emma`, `bf_isabella` |
| `bm_`  | British Male    | `bm_george`, `bm_lewis`  |
| `ef_`  | European Female | varies by language       |
| `em_`  | European Male   | varies by language       |

Browse all available voices in the Voice dropdown inside the studio. Each entry plays a short preview so you can compare styles before committing to a full generation.

## Settings

You can fine-tune Voice Studio behaviour from the controls in the studio and in **⚙️ Settings → TTS Model**:

<CardGroup cols={2}>
  <Card title="Voice & Speed" icon="sliders">
    Select any voice from the library and adjust the speed multiplier. Values below 1.0 slow the speech down; values above 1.0 speed it up. Your last-used voice and speed are remembered between sessions.
  </Card>

  <Card title="MP3 Bitrate" icon="music">
    Choose **128 kbps** for small file sizes, **192 kbps** for a balance of size and quality, or **320 kbps** for the highest-fidelity output. Higher bitrates produce larger files.
  </Card>

  <Card title="TTS Quantization" icon="microchip">
    Found in **⚙️ Settings → TTS Model**. `q4` generates audio faster using less memory. `q8` takes a little longer but produces noticeably cleaner speech — ideal for final deliverables.
  </Card>

  <Card title="Audio Normalization" icon="waveform">
    Audio normalization is on by default to keep volume levels consistent. Disable it in **⚙️ Settings → TTS Model** if you need to preserve the original dynamic range.
  </Card>
</CardGroup>

## Downloading Audio

Once generation completes, the audio player appears with playback controls. Click **Download** to save your MP3. The file is assembled entirely in your browser — no upload, no waiting for a server, no account needed.

<Tip>
  Use **q4 quantization** for quick drafts and listen-throughs — it generates audio significantly faster. Switch to **q8** in **⚙️ Settings → TTS Model** before your final export for the cleanest, highest-quality result. You can toggle between the two at any time without re-entering your text.
</Tip>