Japanese Scientist Breaks New Ground by Translating Mental Images into Text with ‘Mind-Captioning’ AI Technology
November 14, 2025 — A pioneering scientist in Japan has unveiled a novel technology that converts human mental images into descriptive text using a combination of brain scans and artificial intelligence (AI). The breakthrough, detailed in a recent study published in Science Advances, represents a significant step forward in the quest to decode complex visual thoughts—a feat that has eluded researchers until now.
From Brain Activity to Words: The Mind-Captioning Process
Tomoyasu Horikawa, a researcher at NTT’s Communication Science Laboratories near Tokyo, has developed a method called “mind-captioning” that leverages functional magnetic resonance imaging (fMRI) and AI to generate sentences reflecting what people visualize in their minds.
The study involved six native Japanese speakers—four men and two women aged between 22 and 37—who underwent brain scans while watching 2,180 short, silent video clips showcasing a range of objects, scenes, and actions. Using large language models, the AI converted captions corresponding to these videos into numerical sequences. Horikawa then trained simpler AI decoders to link the participants’ brain activity patterns with these numeric representations.
Once trained, the mind-captioning system could analyze real-time brain activity from new videos that were not part of the original training set. An additional AI algorithm progressively generated coherent word sequences that matched the decoded brain signals, resulting in accurate and detailed textual descriptions of the visual content participants experienced or recalled.
Bridging Science and Practical Applications
Unlike previous brain-to-text efforts that primarily focused on translating thoughts involving words, this technology captures richer, more nuanced mental images—including objects, places, events, and their interrelationships. Impressively, the AI model produced output in English despite participants not being native English speakers.
Horikawa points out that the method works even without engaging traditional language-related brain regions, highlighting its potential for helping individuals suffering from language impairments caused by damage to those areas. Practical applications may include aiding people with aphasia or neurodegenerative conditions like amyotrophic lateral sclerosis (ALS), which impair speech production.
Psychologist Scott Barry Kaufman of Barnard College emphasized the profound impact this technology could have for those facing communication challenges, including non-verbal autistic individuals. However, he cautioned that the technology must be used ethically, with strict consent to prevent invasiveness.
Ethical Considerations and Privacy Challenges
While the promise of mind-captioning technology is exciting, experts warn of substantial ethical and privacy concerns. The ability to decode mental images potentially opens the door to revealing private thoughts before individuals wish to verbalize them.
Marcello Ienca, professor of AI ethics and neuroscience at the Technical University of Munich, called this development “the ultimate privacy challenge.” He highlighted the need for stringent regulations to safeguard neural data, especially as companies like Elon Musk’s Neuralink work towards making brain implants available to the public.
A recent Cell journal study suggested methods to prevent unwanted “leakage” of private thoughts by requiring users to think of specific keywords to unlock neural decoding systems, emphasizing the importance of mental privacy protections keeping pace with neuroscience advances.
Łukasz Szoszkiewicz, social scientist at Adam Mickiewicz University and neuro-rights advocate, stressed that neural data should be treated as highly sensitive information by default, advocating for explicit consent, user control, on-device processing, and AI-specific legal frameworks to manage these emerging technologies responsibly.
Limitations and Future Directions
Horikawa acknowledged that his approach still requires extensive data collection and participant cooperation, which limits its current accuracy and practical usability. Moreover, the video clips used depicted common scenarios, and it remains unclear whether the method can accurately interpret more unusual or unpredictable mental images.
Consequently, while the technology might provoke concerns about encroaching on mental privacy, Horikawa reassures that the present method cannot yet “read” a person’s private or spontaneous thoughts without extensive prior data.
Conclusion
The development of “mind-captioning” represents a remarkable leap toward true brain or mind reading, with vast potential for health interventions that could transform communication for many. Continuous scientific innovation paired with rigorous ethical governance will be crucial to harness the benefits of this technology while safeguarding individual freedoms and mental privacy.
For science enthusiasts and curious minds, sign up for CNN’s Wonder Theory newsletter to explore more about astonishing discoveries and technological breakthroughs shaping our world.
