Mistral Voxtral 3 New
Mistral's new speech family: Voxtral TTS 3 text-to-speech with voice cloning and emotion and pace tags, fast multilingual transcription, and a real-time voice agent SDK. Useful for NPC voices, narration and voice-driven game or app features via API.
- Categories
- Voice
- Pricing
- Paid Usage-based API, Voxtral TTS 3 is $13 per 1M characters
- Est. monthly spend
- Light $1/mo · Medium $7/mo · Heavy $30/mo (est.)
Usage-based at $13 per 1M characters; ~0.1M/0.5M/2M+ chars a month. Rough estimates from vendors' public pricing, checked 28 Oct 2026. Not exact — always check the vendor. - Website
- https://mistral.ai/news/voxtral-3/
- Added
- 23 Oct 2026
Compare: Best AI voice generators for game characters and voiceover
Similar tools
ElevenLabs
NewIndustry-leading AI voice synthesis with realistic speech, voice cloning, and music generation. New in Oct 2026: Eleven Realtime speech-to-speech models (v1 Mini and v1 Max, 90+ languages) for low-latency voice agents and interactive characters, via an OpenAI-compatible Realtime API. New Nov 2 2026: SFX v3 sound effects model in the API (eleven_text_to_sound_v3), which follows multi-part prompts more closely and handles vocals and music better.
Est. monthly spend: Free–$22/mo · est.
Google's expressive text-to-speech model in the Gemini API and AI Studio: 2,000+ library voices, voice design from a text prompt, 30-second voice replication, per-line acting direction and two-speaker scenes in 130 languages. New Nov 4 2026: Voice Remixing lets you tweak any voice's timbre, pitch, pace, accent and delivery with a prompt, handy for NPC and character voices.
Est. monthly spend: Unknown
Fish Audio
NewExpressive text-to-speech and voice cloning with emotion tags, powered by its Drama 3 voice model.
Est. monthly spend: Free–$15/mo · est.
Models and inference for consumer AI apps: realtime TTS, STT, speech-to-speech, and LLM hosting.
Est. monthly spend: Unknown