Audio model · from 2 credits
Generate a voice-over with Voxtral TTS 3
Voxtral TTS 3 is Mistral's voice model, released in preview on October 20, 2026. Imaginode offers twelve of its voices: Marie for French, Paul, Jane, Oliver, Sarah, Yael and Eric for English, and one voice each in Spanish, Italian, German, Portuguese and Arabic. You direct the performance in the text, with tags the model applies without saying them: [emotion=calm], [emotion=excited] or [emotion=sad] changes the tone of what follows, [pace=slow] or [pace=fast] the pace. Mistral lists thirty-four emotions. Six thousand characters go through in one block, enough for a five-minute voice-over.
Twelve voices in six languages
Voxtral TTS 3 was released in preview on October 20, 2026. Mistral built it for voice-over and voice assistants, with nine languages announced. On the canvas you choose from twelve voices: Marie for French, Paul, Jane, Oliver, Sarah, Yael and Eric for English, Esmeralda in Spanish, Andrea in Italian, Fiona in German, Marina in Portuguese and Linda in Arabic. Every voice can be previewed before you pick it.
The voices can also read a language other than their own: Sarah reads a French text with a light accent. For a natural French voice-over, stick with Marie.
Directing the performance with tags
Tone is set in the text. An [emotion=…] tag placed before a sentence changes how it is said until the next tag: calm, excited, sad, tense, amused, warm, and about thirty others. A [pace=…] tag sets the speed, from very-slow to very-fast. The two combine, [emotion=calm][pace=slow], and neither is spoken aloud.
In our tests on October 21, 2026, the same sentence went from 3.5 seconds at very-fast to 6.1 seconds at very-slow, and the transcripts of the renders contained no tag. The most common ones are one click away under the text box of the Voice-over node. The price follows the length of the text, tags included, and is shown before you launch.
Voxtral TTS 3 next to its closest siblings
Figures pulled from the same registry as billing: when a price or a cap changes, this table changes with it.
| Voxtral TTS 3 | Fish Audio S1 | Fish Audio S2 Pro | |
|---|---|---|---|
| Base price | 3 credits / 1,000 characters | 3 credits / 1,000 characters | 3 credits / 1,000 characters |
| Voices | 12 | – | – |
| Max prompt | 6,000 characters | – | – |
How to use it on Imaginode
- 1Add a Voice-over node from the toolbar.
- 2Write or wire in the text to read, pick a voice: the price follows the text length.
- 3Generate, listen in the node and download the MP3.
Frequently asked questions
Can I set the speed with the slider?
No, the model ignores that setting. Speed is set with a tag in the text, [pace=slow] or [pace=fast] for example, which also lets you change it partway through.
How long a text does it accept?
Six thousand characters in one block, measured on October 21, 2026: a little over five minutes of voice, rendered in about twelve seconds.
How much does a voice-over generated with Voxtral TTS 3 cost?
3 credits per 1,000 characters, about $0.04 (1 credit = $0.012). The exact cost shows before generating, and the MP3 downloads straight from the node.
Which voices does Voxtral TTS 3 offer?
12 to choose from, with adjustable playback speed, up to 6,000 characters per generation. The voice-over node produces a downloadable MP3, billed per character.
Can I try Voxtral TTS 3 for free?
Yes: signing up grants trial credits, no credit card required. Enough to test Voxtral TTS 3 in real conditions before deciding on a plan or a top-up.
Do I need to install anything to use Voxtral TTS 3?
No: Voxtral TTS 3 runs server-side, in the browser, mobile included. Imaginode's node canvas wires its renders to the other models (references, camera, styles) with nothing to install.
Try Voxtral TTS 3 on a real canvas
129 models, one account, credits that work everywhere. The price always shows before you generate.
Get started
ElevenLabs v3
MiniMax Speech 2.8 HD