Launching today
Voxmint
Natural AI voices in 24 languages, or design your own
2 followers
Natural AI voices in 24 languages, or design your own
2 followers
Text to speech with 113 natural voices in 24 languages. Hanoi and Saigon Vietnamese, Mandarin and Cantonese, Brazilian and European Portuguese. Or describe a voice in a sentence and save it; a fixed seed brings the same voice back in every project. Add [laughing] or [whispering] inline. Batch up to 100 lines, word timings for captions, and a REST API on the same credits. No voice cloning, by design.
Hi Product Hunt, I'm Linh, the solo maker of Voxmint.
I'm in Vietnam, and every TTS tool I tried had the same problem with Asian languages: the free ones sounded robotic or got the tones wrong, and the good ones cost too much for someone making videos every week. So I built a studio around VoxCPM2, an open 2B-parameter model that outputs 48 kHz audio.
There are two ways to get a voice:
• Pick one of 113 catalogue voices. I listened to every one before it went live. That covers northern and southern Vietnamese, Mandarin and Cantonese, and 24 languages in total.
• Describe one. "A warm grandmother telling a bedtime story, slow, a little hoarse." Save it, and a fixed seed brings the same voice back in every episode.
What it doesn't do: voice cloning. That's on purpose. The voice-design filter also refuses real people's names. If you need a specific person's voice, VoxMint isn't the right tool.
For developers there's a REST API on the same credits, with streaming, batches of up to 100 lines and word timings for captions.
One note: this is a new product and has nothing to do with the 2025 "VoxMint" launch on PH.
What I'd love from you: try a line in your own language and tell me if the accent sounds right. Native speakers spot things I can't.