Every Piper voice Quick TTS ships
Piper is the most-selected engine on this site, and it is the one that runs without a GPU. These are the 16 voices you can actually choose, across 13 languages, each with its own page covering the quality tier, what happens when your device cannot load it, and what the alternative on that language is. Everything runs in your browser; nothing you paste is uploaded.
The voices
| Voice | Language | Tier | Role | Step-down? | HD alternative? |
|---|---|---|---|---|---|
ar_JO-kareem-medium |
Arabic | medium |
Default | Yes | Yes |
de_DE-thorsten-medium |
German | medium |
Default | No | Yes |
en_GB-vctk-medium |
English | medium |
Alternative | No | Yes |
en_US-joe-medium |
English | medium |
Alternative | No | Yes |
en_US-libritts_r-medium |
English | medium |
Default | Yes | Yes |
es_ES-davefx-medium |
Spanish | medium |
Default | Yes | Yes |
es_MX-claude-high |
Spanish | high |
Alternative | No | Yes |
fr_FR-siwis-medium |
French | medium |
Default | No | Yes |
it_IT-riccardo-x_low |
Italian | x_low |
Default | No | Yes |
nl_NL-mls-medium |
Dutch | medium |
Default | Yes | Yes |
pl_PL-gosia-medium |
Polish | medium |
Default | Yes | Yes |
pt_BR-faber-medium |
Brazilian Portuguese | medium |
Default | No | Yes |
ru_RU-irina-medium |
Russian | medium |
Default | No | Yes |
tr_TR-fettah-medium |
Turkish | medium |
Default | No | Yes |
vi_VN-vais1000-medium |
Vietnamese | medium |
Default | No | Yes |
zh_CN-huayan-medium |
Mandarin Chinese | medium |
Default | No | No |
By language
- English (English) —
en_US-libritts_r-medium,en_US-joe-medium,en_GB-vctk-medium - Spanish (Espanol) —
es_ES-davefx-medium,es_MX-claude-high· all Spanish engines - French (Francais) —
fr_FR-siwis-medium· all French engines - German (Deutsch) —
de_DE-thorsten-medium· all German engines - Brazilian Portuguese (Portugues) —
pt_BR-faber-medium· all Brazilian Portuguese engines - Italian (Italiano) —
it_IT-riccardo-x_low· all Italian engines - Russian (Русский) —
ru_RU-irina-medium· all Russian engines - Polish (Polski) —
pl_PL-gosia-medium· all Polish engines - Dutch (Nederlands) —
nl_NL-mls-medium· all Dutch engines - Turkish (Türkçe) —
tr_TR-fettah-medium - Mandarin Chinese (中文) —
zh_CN-huayan-medium· all Mandarin Chinese engines - Arabic (العربية) —
ar_JO-kareem-medium· all Arabic engines - Vietnamese (Tiếng Việt) —
vi_VN-vais1000-medium· all Vietnamese engines
The voices you will never see in the dropdown
Quick TTS also carries 5 smaller Piper models that are never selectable: they do not appear in the voice list and cannot be chosen by name. They exist only as the automatic landing place when a device cannot allocate a session for the larger model above them:
en_US-lessac-low— a step-down tier only. It is never user-selectable and never appears in the voice dropdown.es_ES-carlfm-x_low— a step-down tier only. It is never user-selectable and never appears in the voice dropdown.pl_PL-mls_6892-low— a step-down tier only. It is never user-selectable and never appears in the voice dropdown.nl_NL-mls_5809-low— a step-down tier only. It is never user-selectable and never appears in the voice dropdown.ar_JO-kareem-low— a step-down tier only. It is never user-selectable and never appears in the voice dropdown.
They are listed here because the alternative is worse: a reader who sees one of these IDs in a log or an error message should be able to find out what it is, without us implying it is a voice they can pick.
Which one should you use?
On every language except English and Spanish there is exactly one choice, so the real question is which engine to use rather than which voice. Piper is the right answer when you have no usable GPU, and it is the only neural option at all for Mandarin Chinese. Where Supertonic HD covers the language, it is the higher-quality option and it needs WebGPU, which in practice means a desktop.
Speed, volume and voice can all be changed mid-read and the current passage replays under the new settings without starting over. Switching the engine itself stops playback, so you press play once to start the new one. There is no mid-read handover of the remaining text between engines.
Nothing you paste is uploaded
Piper runs as WebAssembly inside your own browser tab. The model is downloaded once and cached; after that, the text you paste is turned into audio on your machine and never sent anywhere. That is an architectural property rather than a policy promise: the test is that generation keeps working with your network disconnected once the model has loaded.