AI Models: Voice, Speech, and TTS Systems | Speech Text

Browse the AI voice and speech models we run and write about. Play with the interactive demos, then open a dedicated guide for any of them. We list new models here as they ship.
Every model gets its own page: a live demo, a rundown of what it can do, and a straight, hands-on guide.
NVIDIA's open 0.6B multilingual, streaming speech-recognition model. Run Nemotron AI audio-to-text live: drop in a clip or a URL and read a punctuated transcript.
ByteDance's Seed Audio 1.0, served on fal, can generate audio from a prompt, reference audio or an image. Set up the API request and read the responsible-use notes.
MisoTTS, the open-weights, richly emotive text-to-speech model from Miso Labs. Open the live demo, clone a voice from a short clip, and listen to lifelike speech.
A multilingual, zero-shot TTS model from k2-fsa covering 600+ languages. Explore the live Hugging Face demo for voice cloning, voice design, and natural AI speech.
VoxCPM2 from OpenBMB: a tokenizer-free TTS model handling 30 languages, voice design, controllable voice cloning, and 48 kHz speech. Open the official Hugging Face demo.
More AI voice and speech models are on the way. Check back soon, or dive into Nemotron 3.5 ASR, Seed Audio 1.0, Miso One, OmniVoice, or VoxCPM right now.
Each of these works hand in hand with our core speech-to-text workspace.
Generate lifelike, expressive speech and clone voices using the open MisoTTS model.
Upload or record audio to get accurate AI transcription, captions, and notes powered by Whisper.
Speech Text is an online speech to text workspace powered by OpenAI Whisper technology. Convert audio, video, meetings, interviews, podcasts, lectures, and caption sources into searchable, editable text.
Speech Text © 2026. Built for online transcription.