Block develops Dessa, a speech-to-speech model trained only on human voice
Article timeline
The articles in this story. Times are ET.
- MarkTechPostNewsBlock releases Dessa, a full-duplex speech model trained only on speech
AIBlock's AI research lab announces Dessa, a speech-to-speech model that listens and talks at the same time, trained only on human conversation audio with no text or pretrained models. Block says it trained Dessa on one NVIDIA GB200 NVL72 rack, compared with 1,016 H100 GPUs used for Kyutai's Moshi, and that it runs locally on a MacBook Pro M4 Max faster than real time. The research preview has no published benchmarks, parameter count, weights, API, or price, and access is by waitlist.
meng shaoXPickBlock launches Dessa, a full-duplex speech model trained only on human voiceAIBlock says Dessa is a speech-to-speech model trained from scratch on human voice, with no text or pretrained base, and it claims inference faster than real time. The model listens and speaks at the same time, and it can run locally on a MacBook Pro M4 Max or better. Block argues that converting speech to text loses tone and emotion, which it says is both a user-experience and a safety problem.
- IThome · AINewsBlock unveils Dessa, a full-duplex voice model that runs locally on MacBooks
AIBlock, the company co-founded by Jack Dorsey, has released Dessa, a full-duplex voice model that can express laughter, pauses and choking. The company says Dessa was trained only on human speech data using a single rack of Nvidia GPUs and can run locally on consumer MacBooks with M4 Max or higher. Dessa is not yet open for public trial, and Block is inviting safety researchers to test risks such as voice agents deceiving humans.
- TechCrunch · AINewsPickBlock releases Dessa, a full-duplex voice model that runs locally
AIBlock's research lab released Dessa, a full-duplex speech-to-speech model that can talk and listen at once and runs on consumer hardware such as a MacBook M4 Max. Block says emotion is built into the model rather than driven by text or voice tags, but the source does not say whether tags are used. Dessa is not publicly available yet, and Block has asked safety researchers to test it for ways a voice agent could deceive people.
- The DecoderNewsBlock unveils Dessa, a speech AI model trained only on human speech
AIBlock has unveiled Dessa, an AI speech model that it says is the first trained entirely on human speech with no text data. Dessa processes audio directly rather than converting it to text, which lets it detect emotions, sarcasm, laughter, and pauses, and it supports full-duplex conversation. Dessa is not publicly available; Block is first recruiting AI security researchers to test it for risks.
- Gizmodo · AINewsBlock's Dessa speech model trained only on human conversation
AIBlock released Dessa, a speech-to-speech AI model trained from scratch on human conversation, which processes and generates audio directly rather than converting speech to text. The model can speak and listen at the same time, and Block says it can run locally on any MacBook Pro M4 Max or better. Block has given early access to a small group of AI safety researchers.
🚨 AI News | TestingCatalogXJack Dorsey announces Dessa, a full-duplex speech model trained only on voiceAIJack Dorsey announces Dessa, a full-duplex speech model trained solely on voice conversations, with no text data or pretrained models. Only demo samples are available so far, which the post says far surpass other models in emotional expression and natural feel.
- The Verge · AINewsBlock unveils Dessa, a speech-to-speech model trained only on human conversation
AIBlock has announced Dessa, a speech-to-speech AI model that it says was trained from scratch only on human conversation. Block shared sample chats that sometimes sound awkward, and the Dessa team says it can run locally on a MacBook Pro M4 Max or better. The team plans to give early access to a small group of AI safety researchers.