AI Research 73% 1 min readJul 6, 2026, 3:14 PM

Kyutai's Pocket TTS clones a voice from 5 seconds of audio, on CPU, under MIT. Benchmarked against Kokoro, Supertonic, and Inflect-Nano for Eng. TTS

30-second summary

Kyutai's Pocket TTS can clone a voice from just 5 seconds of audio on a CPU, and has been benchmarked against other models. It achieved notable results despite being the slowest model tested.

Kyutai's Pocket TTS clones a voice from 5 seconds of audio, on CPU, under MIT. Benchmarked against Kokoro, Supertonic, and Inflect-Nano for Eng. TTS
Key takeaways
  • Pocket TTS can clone a voice from just 5 seconds of audio
  • The model runs on CPU, making it more accessible to users
  • Pocket TTS was benchmarked against other notable TTS models, including Kokoro and Supertonic
  • The model achieved notable results despite being the slowest tested
Full story

Kyutai's Pocket TTS is a text-to-speech model that can clone a voice from a very short audio sample, just 5 seconds, and run on a CPU. This is a significant achievement in the field of AI-powered voice synthesis.

The model was benchmarked against other notable CPU-based TTS models, including Kokoro, Supertonic, and Inflect-Nano. The benchmarking process involved 180 timed runs and 36 audio samples, with objective MOS scores provided via UTMOS.

The results showed that while Pocket TTS is the slowest of the models tested, it is still a highly interesting and capable model. Its ability to clone a voice from such a short audio sample is a major advantage, and its performance on CPU makes it accessible to a wider range of users.

The implications of this technology are significant, with potential applications in areas such as voice assistants, audio production, and accessibility tools.

Source: Kyutai's Pocket TTS clones a voice from 5 seconds of audio, on CPU, under MIT. Benchmarked against Kokoro, Supertonic, and Inflect-Nano for Eng. TTS. Read the full piece at the source.

Why this matters
Developers

offers a new tool for voice synthesis and audio production

Businesses

potential applications in voice assistants and customer service

Everyone

advances in AI-powered voice synthesis

Glossary
MOS scores
Mean Opinion Score, a measure of audio quality
UTMOS
a tool for evaluating the quality of text-to-speech models
Sources ยท 1
Related
TickrWire

AI news intelligence. We aggregate, verify, summarise and explain the latest artificial intelligence news from open, legal sources.

Daily AI digest

Top AI stories, summarised, in your inbox each morning.

ยฉ 2026 TickrWire. Summaries and analysis are AI-generated and may contain errors.Privacy