Maceió-based Hand Talk's AI avatar Hugo translates Portuguese into Brazilian Sign Language and has been downloaded over 2 million times, citing 3 billion translated words, though Deaf users report it misses sign-language grammar and cannot replace human interpreters.
2 million+
App downloads
3 billion
Cumulative words translated
~15,000 entries, 100+ countries
UN World Summit Award competing entries (2013)
Details
Maturity
Established
Promoter
Hand Talk (acquired by Sorenson, 2025)
Period
2012–ongoing (acquired by Sorenson, Jan 2025)
Keywords
sign language translation, Libras, deaf accessibility, avatar AI, assistive technology
Context
Hand Talk is a Brazilian company founded in 2012 in Maceio, whose mobile app and browser widget use an AI-driven 3D avatar, Hugo, to translate Portuguese (and more recently English) into Brazilian Sign Language (Libras) and American Sign Language. In schools it is used as a supplementary tool -- a Libras dictionary for academic vocabulary and an on-demand translator for when a human interpreter is unavailable.
Activities
The app won the United Nations World Summit Award for Best Social Application in 2013, competing against roughly 15,000 entries from over 100 countries. Brazilian broadcasters' accessibility requirements helped drive adoption, and in January 2025 Hand Talk was acquired by Sorenson, a global accessibility-technology company that also owns the automated sign-language translator OmniBridge.
Results
Hand Talk has been downloaded more than 2 million times, with the company citing 3 billion words translated cumulatively. Deaf users' reviews describe real limitations: several report that Hugo defaults to finger-spelling words for which it has no stored sign, misses Libras/ASL grammar and word order, and can feel more like a novelty than a substitute for a human interpreter, with at least one account describing a Deaf family member feeling diminished rather than helped. No independently published accuracy or bias evaluation of Hugo's sign output was found, and Hand Talk does not publish classroom learning-outcome data.
Implementation
Indicative cost
Low (< €50k)
Time to results
Short (< 1 year)
Conditions for success
Used as a supplementary classroom tool alongside, not instead of, human interpreters
Broadcast-accessibility requirements in Brazil helped drive real-world adoption
Common failure modes
Hugo reportedly defaults to finger-spelling words it has no stored sign for
Misses Libras/ASL grammar and word order according to Deaf user reviews
No independently published accuracy or bias evaluation of the sign output exists
Where it fits
Governance type
commercial/private (education used as supplementary tool)
Scale
national (Brazil) and international
Income level
upper-middle-income
Data sources
Where this practice's information was retrieved from, and when.
A Nairobi startup's 3D-printed, recycled-plastic humanoid robots translate speech to sign language in real time for deaf students in low-connectivity …
STMicroelectronics Foundation, ITU and a blind Milan professor run train-the-trainer digital-literacy classes for blind and visually-impaired adults in Dakar. An …