Best free Kyutai TTS & STT alternatives

✦ WhyA Kyutai TTS & STT alternative — Only 82M parameters yet ranks near top of the TTS Arena leaderboard for quality.

• Only 82M parameters yet ranks near top of the TTS Arena leaderboard for quality
• Apache 2.0 license — fully free for commercial use, weights open on Hugging Face
• Trained on a shoestring budget, the cheapest-to-reproduce competitive TTS model

✦ WhyA Kyutai TTS & STT alternative — MIT-licensed, fully open-source TTS/voice-cloning family (Turbo/Nano/Multilingual V3).

• MIT-licensed, fully open-source TTS/voice-cloning family (Turbo/Nano/Multilingual V3)
• Built-in watermarking for responsible-AI provenance on generated audio
• Paralinguistic tags ([laugh], [cough]) for expressive, non-flat speech

✦ WhyA Kyutai TTS & STT alternative — Turns text into expressive speech using Microsoft's MAI-Voice models.

• Turns text into expressive speech using Microsoft's MAI-Voice models
• Control emotion, style and personality
• Free in the browser, no mandatory login per press reports
• MAI-Voice-2 supports 15 languages

✦ WhyA Kyutai TTS & STT alternative — Open-source TTS generating up to 90 minutes of speech with up to 4 speakers.

• Open-source TTS generating up to 90 minutes of speech with up to 4 speakers
• VibeVoice-ASR transcribes 60-minute audio in one pass with speaker, timestamps and text
• ASR covers 50+ languages
• Free to self-host from GitHub and Hugging Face

✦ WhyA Kyutai TTS & STT alternative — Multilingual voice generation with zero-shot voice cloning.

• Multilingual voice generation with zero-shot voice cloning
• Streaming synthesis with low latency
• Inference and training code plus weights
• Apache-2.0 open source

✦ WhyA Kyutai TTS & STT alternative — Open Apache-2.0 conversational speech model generating context-aware, natural speech.

• Open Apache-2.0 conversational speech model generating context-aware, natural speech
• Takes prior dialogue turns as context for prosody
• Weights on Hugging Face; base for the Sesame voice companion

✦ WhyA Kyutai TTS & STT alternative — Llama-based speech LLM with human-like intonation and emotion tags.

• Llama-based speech LLM with human-like intonation and emotion tags
• Zero-shot voice cloning and low-latency streaming
• Apache-2.0 with multilingual variants

✦ WhyA Kyutai TTS & STT alternative — Flow-matching TTS with zero-shot voice cloning from a few seconds of audio.

• Flow-matching TTS with zero-shot voice cloning from a few seconds of audio
• English, Chinese and community multilingual checkpoints
• MIT code, free Hugging Face Space and Gradio app

✦ WhyA Kyutai TTS & STT alternative — Bilibili's zero-shot TTS with precise duration control and emotion control (IndexTTS2).

• Bilibili's zero-shot TTS with precise duration control and emotion control (IndexTTS2)
• Strong Chinese pinyin correction plus English
• Open weights and demo on Hugging Face