Block Unveils Dessa, a Full-Duplex Speech Model Trained Only on Speech
Block's AI lab released Dessa, a full-duplex speech-to-speech model trained only on human conversation audio. It runs locally on a MacBook Pro M4 Max, but no size, benchmarks, or weights were published.
Key Takeaways
- Key Highlight：Block's AI lab released Dessa, a full-duplex speech-to-speech model trained only on human conversation audio. It runs locally on a MacBook Pro M4 Max, but no size, benchmarks, or weights were published.
- Key Topics：Block, Unveils, Full-Duplex, Speech
Block's AI lab has introduced Dessa, a full-duplex speech-to-speech model trained exclusively on human conversation audio. Unlike many speech systems that rely on text transcripts or multimodal data, Dessa learns directly from raw speech, aiming to capture natural turn-taking and interruptions.
The model runs locally on a MacBook Pro M4 Max, suggesting a focus on on-device efficiency and low-latency interaction. Full-duplex capability means it can listen and speak simultaneously, which is a key challenge for conversational AI. However, Block has not disclosed the model's size, benchmarks, or weights, limiting independent verification.
If Dessa performs well, it could push forward real-time voice assistants and on-device AI. But without public benchmarks or open weights, its practical impact remains unclear. It may serve more as a research showcase than a deployable product, though it signals continued interest in speech-only training approaches.
This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.