BCIwiki (bciwiki.com) — Researchers have built Brain2Speech-Net, a single-stage framework that generates speech from neural signals without first decoding them into text, according to a preprint posted to arXiv on September 3, 2026. The authors report it was the only system in their comparison to produce intelligible speech while generating faster than real time. The work has not been peer reviewed.
Loss of speech limits communication for people with paralysis, the authors note, and the scarcity of neural data for training speech brain-computer interfaces (BCIs) makes direct neural-to-speech synthesis difficult. Most existing systems use cascaded neural-to-text-to-speech pipelines, which raise inference latency and let errors carry over from one stage to the next.
Brain2Speech-Net uses a differentiable phoneme bottleneck and a deep hidden Markov model (deep-HMM) alignment mechanism to map long neural recordings into the latent space of a text-to-speech (TTS) model, which the authors say enables high-quality speech synthesis.
The authors are Shreeram Suresh Chandra, Zexin Cai, Yu Tsao, Simon King and Berrak Sisman. The abstract does not specify the recording modality, the number of participants or intelligibility scores.