Voice AI

Specialists from NVIDIA have introduced Nemotron 3 Diarization — an advanced system for speech recognition. It is capable of analyzing audio tracks with multiple speakers, even if they are arguing, and accurately determining who is speaking and at what moment. The model supports dialogues involving up to eight people, which is the best result compared to analogues. At the same time, Nemotron 3 Di…

ACE-Step 1.5 XL has been released, an open-source audio model that outperforms competitors in benchmarks and can even run on low-spec computers. Capabilities: • Generates music, sounds, and audio for any task; • Delivers high detail even in complex musical compositions; • Creates a full track in just 4–8 steps with minimal adjustments; • Produces four-minute compositions in seconds, supporting du…

In just a few seconds, you can create a copy of a voice—a new simplified model for speech synthesis has emerged, capable of reproducing someone else's voice from a short audio clip. It's all very simple: just provide the neural network with a few seconds of a person's recording, and it can speak any text in that same voice. The result sounds quite natural: the quality reaches 48 kHz, which is com…

A powerful open-source speech generator has been released — the Tada model was just introduced but is already positioning itself as a competitor to leading services. Capabilities: • Easily handles long texts — supports up to 700 seconds of audio and 2048 tokens; • Generates highly accurate speech without unnecessary insertions; • Operates approximately five times faster than many similar programs…