Table of Contents
Universal Motion Intelligence has launched a Bittensor-based AI network for motion-to-meaning intelligence, beginning with /sign, a product designed to support real-time communication between Deaf and hearing people. The product is being built to translate signed communication into spoken language and spoken replies into captions and signing, using a standard camera rather than gloves, wearables, or specialized sensors.
The company announced the launch as a repositioning of Subnet 78, the Bittensor subnet formerly known as Vocence. Instead of focusing only on voice AI, UMI is now orienting SN78 around models that interpret motion, gesture, expression, timing, and physical context.
Koyuki Nakamori, UMI co-founder and Head of AI and former Head of AI at the OpenTensor Foundation, said sign language is a difficult AI problem because meaning is carried across hands, body movement, facial expression, and timing at the same time. UMI is using Bittensor to turn that challenge into an open model competition instead of relying only on a single closed system.
/sign Targets Real-Time Deaf–Hearing Communication
/sign is designed as a two-way communication tool. In one direction, a camera captures a Deaf user signing, the system interprets the movement, and the response is rendered as speech. In the other direction, spoken language is transcribed, translated, captioned, and represented visually for the Deaf user.
The underlying stack is a layered engine that combines 3D visual capture, skeletal and gesture modeling, facial-expression analysis, contextual language understanding, multilingual voice, and live inference. UMI says the first language pair in active development is ASL to English, with the broader architecture aimed at supporting more sign languages over time.
With /sign, the team is targeting under three seconds for end-to-end translation latency (verified benchmarks will be published later in a technical whitepaper). The launch introduces the product direction and network model, while the most important performance claims still depend on public benchmarks.
UMI also says /sign is not intended to replace certified interpreters where one is required or appropriate. Instead, /sign is for everyday and on-demand situations where an interpreter is not available, including commercial, healthcare, and public-service settings where accessible communication requirements already exist.
How Bittensor Fits Into the UMI Network
The Bittensor component is the competitive evaluation layer behind UMI’s motion-intelligence work. Miners can submit models, while validators evaluate outputs against defined benchmarks. The setup is a King of the Hill system where the best-performing models rise through public scoring.
That structure is meant to make SN78 an open market for motion-to-meaning models instead of a conventional software product controlled by one vendor.
If the system works as intended, independent teams can compete on translation quality and other motion-understanding tasks, while validators provide the performance feedback that determines which models receive weight.
The subnet is already live on mainnet, with miner participation and a public miner pilot open. Official rankings have not started.
A Broader Motion-Intelligence Roadmap
UMI is not presenting /sign as the full extent of the subnet.
The same motion-to-meaning approach could support a broader family of applications, including /face for facial understanding, /intent for body language and human intent, /flow for crowd movement, /move for embodied AI, /motion for human activity recognition, and /trust for behavioral credibility.
The company also plans developer infrastructure through /sdk and /api, which would make the underlying capabilities available beyond the initial consumer-facing product. That broader roadmap puts UMI closer to a general physical-intelligence layer than a single sign-language app, though /sign is the first application with a public launch timeline.
Early access requests, clinical pilot inquiries, and launch partnerships are open through umi.vision. UMI says the web app is rolling out in the coming week, TestFlight beta access opens October 15, and the App Store launch is planned for November 15.
