Clevr
Finally, an AI that talks and explains it visually
Start new thread

Clevr Labs - Banter 1: Human-like text-to-speech for Arabic & English

by•
Banter 1 is Clevr Labs' first speech model: real-time, natural-sounding text-to-speech built for conversation, not narration. It captures the disfluencies, backchanneling, and vocal bursts that make speech feel human, with under 200ms latency. Supports English and Arabic, and is built for voice agents, conversational AI etc.

Add a comment

Replies

Best
Maker
📌
Hey PH, I'm Mustafa, one of the co-founders of Clevr Labs. Today we're launching Banter 1 - our first Text to speech model, trained end to end by us. Quick context on who's behind it: we're two 22-year-olds in Dubai who taught ourselves ML over the past few years. Banter 1 is real-time text-to-speech built for conversation, not narration. It captures the disfluencies, backchanneling, and vocal bursts that make speech feel human, with under 200ms latency. Supports English and Arabic, and is built for voice agents, conversational AI etc. - ~136ms to first audio, consistently under 200ms under load - $2.40 per hour of generated audio - Straightforward TTS API — drop it into whatever you're building Try the demo, NO SIGNUP. We wired Banter 1 into a voice agent so you can just talk to it and hear it answer. Thanks :)