Whats the best open source/free TTS

Posted by NightMatko@reddit | LocalLLaMA | View on Reddit | 12 comments

Hey, Im trying to see how much does synthetic data help with training ASR model. What is the best TTS? Im looking for something that sounds natural and not robotic. It would be really nice if the TTS could mimic english accents (american, british, french etc.). Thanks for the help.

[-]

mvdirty@reddit

For me, at least, Qwen3-TTS is still beating the others folks have been mentioning so far, for both speed and quality of voice-cloned generation. Use its voice design or built-in voices if you want emotional control, or use its voice cloning with your favorite acquired recordings and vary emotion by having a small selection of reference audio files you choose from. You'll have no issue with accents if you use its voice cloning, that much I can promise you.

[-]

Ordinary_Lemon_5238@reddit

im trying to use qwen3-tts in pinokio and the voice clone and design tabs just freeze when i try to click them, any idea why? how do you run Qwen?

[-]

mvdirty@reddit

Sorry for replying so late; I've been away from Reddit pretty much all of the last week. I use Qwen3-TTS via mlx-audio, so I'm afraid I can't be of much help to you. I hope you've been able to get things sorted out in the time I've been away!

[-]

Novel_Leading_7541@reddit

Use open-source TTS carefully—some models aren’t commercial-friendly (e.g., Fish Audio and Voxtral use CC BY-NC 4.0, which prohibits commercial use). For overall quality and realism right now, Qwen3-TTS is one of the strongest options, especially for natural speech and accent flexibility.

[-]