When you think a serious AI project needs a massive team, lots of money, and years of experience – two students from Korea show up and prove otherwise. The team from Nari Labs, a small startup founded by Toby Kim and his colleague, has just released Dia, a new open-source text-to-speech model.
And not just any model – this one already outperforms some of the most well-known commercial tools like ElevenLabs and Sesame.
The model has 1.6 billion parameters and supports features most others still don’t, like emotional expression, speaker recognition, and even non-verbal sounds like laughter, coughing, or screaming. In short, it sounds like a real person – but even more natural and lively.
What’s especially impressive is how they pulled it off. Inspired by Google’s NotebookLM tools and with free access to Google’s TPU Research Cloud, they managed to train an advanced model with zero budget – and in direct tests, it shows better results than major company products in speed, expressiveness, and understanding the “unsaid.”
And they’re not stopping there.
Nari Labs is already developing an app for end users – something that would let everyday people create and remix voice content with ease.
In Brief: Tech World Highlights
- Nissan has announced a partnership with British AI startup Wayve to integrate Wayve’s autonomous driving technology into its vehicles.
- Amazon CEO Andy Jassy published his annual letter to shareholders, stating that generative AI “will reshape nearly every customer experience we know.”
- Hugging Face has acquired Pollen Robotics and introduced Reachy 2, a $70,000 humanoid robot designed for research and applications involving integrated AI technology.
- IFS, a Swedish provider of cloud software and industrial AI technology, has reached a valuation of over €15 billion ($17 billion) due to rising demand.
- Third-party testing and internal evaluations revealed that OpenAI’s new o3 and o4-mini models exhibit significantly more hallucinations compared to older models.
Trending AI Tools:
- ChatGPT – New memory feature that remembers all previous conversations.
- Grok 3 – xAI’s flagship model, now with advanced memory capabilities.
- Canva Visual Suite 2.0 – Design creation powered by AI across multiple formats.
