Docs / Guides
Ways to Read Aloud: Cloud, Offline Voice Packs, Local CosyVoice, System Voices
Soniva desktop has five ways to read text aloud. They differ in three things: cost, whether you need the internet, and whether the audio can be saved to your library.
Soniva desktop has five ways to read text aloud. They differ in three things: cost, whether you need the internet, and whether the audio can be saved to your library.
Applies to Soniva desktop (Mac / Windows). Checked on 2026-10-14.
1. At a glance
| Method | Cost | Internet | Saved to library automatically | Good for |
|---|---|---|---|---|
| Cloud AI voices (Soniva credits) | Subscription + credits | Required | ✅ | Best quality, no setup |
| Cloud AI voices (your own API key) | Billed by the provider, no credits | Required | ✅ | You already have a provider account |
| Offline voice pack (Kokoro) | Free | Only to download the pack | ✅ | Saving money, offline use, quality is good enough |
| Local CosyVoice | Free | Only to install | ✅ | Long documents on a Mac, voice cloning |
| System voices | Free | Not needed | ❌ | Occasional listening |
The two cloud options are covered in Voice engines; this page focuses on the other three.
2. Offline voice pack (Kokoro)
Speech is generated on your own computer, without any server.
- How to turn it on: Settings → Text to speech → choose "Local voice model" as the brand → download a voice pack. The download starts right away; no need to pick files on a website.
- Three editions (same voices — one is enough):
| Edition | Download | Memory while reading |
|---|---|---|
| Light | ~160 MB | ~450 MB |
| Standard | ~207 MB | ~700 MB |
| High quality | ~361 MB | ~1.3 GB |
- Voices: Chinese and English, male and female.
- Saved to library: each segment is saved as audio while it plays, tagged "Kokoro" plus the voice name; reading the same text with the same voice again doesn't save a duplicate.
- Commercial use: Kokoro is released under Apache-2.0, so the generated audio can be used freely, including commercially.
- Memory: if your computer is low on free memory, Soniva asks before generating.
3. Local CosyVoice (Mac only)
Alibaba's open-source CosyVoice model, running on your own Mac.
- How to turn it on: Settings → Text to speech → choose "Local voice model" as the brand → pick CosyVoice under "Model" on the left, then click "Install". Installing, starting and stopping are all buttons in the app — no command line.
- Resources: about 4–8 GB of memory while running; the service stops when you quit Soniva.
- Once enabled: all reading goes through it (exclusive mode). If the service isn't running you get a clear message — Soniva never silently switches to another voice.
- Good for: reading long documents in bulk. For live chat, a lighter voice is better.
- Voice cloning: you can clone a voice from a recording. Only clone your own voice, or a voice whose owner has agreed.
- Saved to library: each segment is saved, tagged "CosyVoice" plus the voice name.
- Commercial use: CosyVoice is open-source under Apache-2.0 and the generated audio can be used freely; cloning someone else's voice needs their permission.
- Windows: not supported yet.
4. System voices
Uses the reading engine built into Mac or Windows.
- Free, no sign-in, no limits.
- Not saved to your library, for two reasons:
- The system engine only plays the sound; it doesn't hand the audio to apps, so Soniva has no audio file to save;
- System voice licenses are restricted. Apple, for example, only allows system voices for personal, non-commercial use, not recording them for publishing or commercial use.
- For free audio you can save and use commercially, use the Kokoro voice pack above.
5. Which one to pick
| Situation | Suggestion |
|---|---|
| Occasional listening, no need to save | System voices |
| Don't want to pay, but want audio files | Kokoro voice pack |
| A Mac with plenty of memory, lots of long text | Local CosyVoice |
| Best quality or many languages | Cloud AI voices |
Related
Was this article helpful?
Anonymous feedback is enabled when you allow analytics and the service is available. You can also report an issue by email.
The email includes this article. You can also copy the details and send them to soniva.cde@gmail.com