How to Use Read Aloud with ElevenLabs or the System Voice
Use Read Aloud with ElevenLabs or the system voice in Snipset: your own API key, voices and models, the 1.2x speed limit, and how fallbacks work.
Read Aloud in Snipset turns the text on your clipboard into speech, with words stepping across an overlay as they are spoken. You can use Read Aloud with ElevenLabs, through your own API key, or with the system voice that already lives on your computer. This tutorial walks through both on Windows and Linux: how to switch engines, add your key, pick a voice and model, and what to expect from speed and fallbacks.
What Read Aloud Does
Copy text in any app, then press the Read Aloud shortcut (Ctrl+Alt+R by default). Snipset reads the clipboard, not the on-screen selection, so a highlight you never copied is not read. An overlay shows the words as they are spoken, with controls to pause, replay, and stop.
You change the key under Settings → Audio & Voice → Read Aloud & RSVP Shortcut, and the pace under Read Aloud Speed (WPM), which starts at 250 words per minute.
The System Voice: What Speaks on Your Computer
The default engine is System voice. It plays through the Web Speech API of the app's WebView, so it uses voices your operating system already provides:
- Windows: WebView2 offers the voices installed in Windows.
- Linux: WebKitGTK asks
speech-dispatcherfor voices. If the voice list is empty, installspeech-dispatcherandespeak-ng, check withspd-say "Hello", then restart Snipset. espeak-ng is clear but robotic; it is a dependable offline voice, not a natural-sounding one.
The system voice needs no account and no key.
How to Use Read Aloud with ElevenLabs
If you want a more natural voice, ElevenLabs is the second engine. You bring your own ElevenLabs account and API key; Snipset sends the text straight to the ElevenLabs API (api.elevenlabs.io). ElevenLabs bills that usage per character on your own account, so check your plan in the ElevenLabs dashboard.
Add your API key
- Open Settings → Audio & Voice and find Text to speech engine.
- Set Voice engine to ElevenLabs.
- Paste your key into ElevenLabs API key and press Save. Show key lets you check what you pasted, and Remove deletes it.
- Press Test to confirm the key works.
If you create a restricted key in ElevenLabs, give it both the Text to Speech and Voices (read) permissions. Without the second one, Test tells you: "This key needs the Text to Speech and Voices (read) permissions." You do not need a new key; add the permission to the existing one and test again.
Pick a voice and a model
Once the key works, the Voice list fills with the voices in your ElevenLabs account. Press Reload voices after you add new voices on the ElevenLabs side. For Model, there are two choices:
| Model | When to pick it |
|---|---|
| Flash v2.5: faster | Quick replies and everyday reading |
| Multilingual v2: better with numbers | Text full of figures, dates, and prices |
Where the key is stored
The key is not kept as plain text in Snipset's settings. On Windows it is encrypted with your Windows account (DPAPI). On Linux it goes into your system keychain through Secret Service, which GNOME Keyring, KeePassXC, or KDE Wallet can provide. If no keychain is available on Linux, Snipset asks you to install or unlock GNOME Keyring or KeePassXC, or to turn on the Secret Service option in KDE Wallet.
Speed: Why ElevenLabs Tops Out at 1.2x
Your Read Aloud Speed (WPM) setting is turned into a speaking rate against a baseline of 150 words per minute. The default 250 WPM asks for about 1.67x. The system voice follows that rate.
ElevenLabs works differently. Its documentation says voice speed ranges from 0.7 to 1.2, so Snipset keeps every request inside that range. A faster setting is read at 1.2x. The overlay follows the voice you actually hear: with ElevenLabs it steps at the voice's real pace (at most 1.2 times 150, so 180 words per minute) instead of racing ahead at the WPM you chose. With the system voice, the overlay keeps your chosen WPM.
Early builds sent a value just outside that range at fast settings, so ElevenLabs refused it and the local voice read instead. That was fixed on Linux in 0.6.877 and on Windows in 0.6.859.
When ElevenLabs Is Unavailable
ElevenLabs can fail for ordinary reasons: a key without the right permission, used-up credits, a voice your plan cannot use, no voice picked, or no connection. Read Aloud then falls back to a local voice on your device, so keep at least one voice installed.
While desktop notifications are on, Snipset also shows a notification that says why, for example "Your ElevenLabs credits are used up, so Read Aloud used the local voice." You see it once per reason until Snipset restarts, so a long reading session does not flood you. Every fallback, not just the first, is written to the log with its reason, which makes "why did it sound robotic just now?" easy to answer later.
Read Aloud on Wayland
On Linux with Wayland, global shortcuts go through the desktop portal. Before linux 0.6.876, the portal sent the Read Aloud key to the recording toggle, so pressing it opened the microphone and nothing was spoken. Now the portal activation for Read Aloud starts reading on every press, the same way the X11 hotkey does. Dictation shortcuts on Wayland still toggle on each press, because the portal does not report key releases.
Try Both Voices
Start with System voice to get a feel for Read Aloud with nothing to set up, then add an ElevenLabs key when you want a more natural voice for long documents. Switching takes one setting, and the fallback keeps reading if ElevenLabs is unreachable. The full reference is in the audio documentation.
Read Aloud with both engines is available in Snipset for Windows and Linux. See Snipset plans and pricing →