9:00 AM The bank
A question about a loan. The interpreter was booked for next week. Today it is a pen, a notepad, and a lot of guessing on both sides of the counter.
Real-time sign language
You speak. The avatar signs.
SignaNova turns speech and text into sign language as you speak, performed by a photorealistic avatar built from the movement of Deaf native signers. Working today in American Sign Language. Croatian Sign Language next.
A conversation should not need an appointment.
Imagine an ordinary Tuesday.
For a Deaf person, three ordinary errands. Each one depends on someone else being available to translate.
11:30 AM The doctor
Symptoms, dosage, side effects. A family member interprets, so privacy is gone and the details are approximate.
5:00 PM The school
A parent meeting about their child. Twenty minutes of nodding along, and the real conversation never happens.
None of this is dramatic. That is the point. It is every day.
70M+
Deaf people worldwide; for most, a sign language is the first language and written text a second one.
134
recognized sign languages, each a full language with its own grammar, not a gesture version of speech.
Scarce
interpreters: booked days ahead, rarely there when life actually happens.
A conversation that simply happens.
You speak. The avatar signs.
Working todayA photorealistic avatar signs what you said in the grammar of the person's own sign language, as you say it. American Sign Language first.
They sign. You hear.
In developmentA standard camera reads the signing and gives you their words as text or voice. Same conversation, both directions.
What changes for the person on the other side
Spontaneous
No booking, no waiting. The conversation starts when you walk in.
Private
No relative in the room, no stranger hearing your diagnosis.
Dignified
You are talked to, not talked about. Eye to eye.
Why it moves like a person. Because it is one.
Every sign the avatar performs was first performed by a Deaf native signer in our own motion capture studio. Hands, fingers, face, posture, rhythm, all recorded in a single take. The avatar does not imitate signing. It replays a human performance, preserved in three dimensions.
- Deaf native performers, not animators guessing.
- Our own in-house Vicon Valkyrie system: full body, ten fingers and face in one take.
- Facial expression captured in the same take, never added later.
- A photorealistic MetaHuman avatar, rendered in real time in Unreal Engine 5.
How it works, in one line.
A short chain. Each link is simple to say and hard to build.
-
Capture
Deaf performers recorded in our studio, body, fingers and face in one take.
-
Validate
Deaf native signers review every sign before it enters the library.
-
Translate
Speech becomes text, text becomes sign language grammar, grammar becomes the right sequence of signs.
-
Blend
Signs are joined with natural transitions and synchronized facial expression, so it reads as one performance.
-
Render
A photorealistic avatar in Unreal Engine 5, in real time, on the device in your hand.
When a word has no sign in the library yet, such as a name or a rare term, the avatar fingerspells it, as a human signer would.
Built with the Deaf community, not for it.
Sign languages are living languages with regional variation, grammar and nuance that no hearing engineer can judge. So we do not.
Deaf native performers
Every sign is performed by people for whom that sign language is their first language.
The Deaf community validates
Grammar, timing, facial markers and regional variants are checked by Deaf native signers under a documented validation protocol before anything is used.
Nothing ships unvalidated
If the community says a sign is wrong, it is wrong. It goes back to the studio. There is no override.
Partnered from day one
Developed in partnership with the national Deaf association in our home market.
Accuracy here is not a metric. It is respect.
Ethical AI, built in from day one.
SignaNova is an AI interpreter for routine, everyday interactions, standing beside a human skill that carries people's health, money and rights. Our rules are written down, and we hold ourselves to them.
It says it is an AI
The avatar is always labeled as AI-generated, is never presented as a human interpreter, and is not a copy of any real person.
Humans stay essential
Court testimony, therapy, end-of-life decisions, high-stakes negotiation belong to certified interpreters; we write that boundary into every deployment contract.
Deaf people hold the veto
Every sign is performed by a Deaf signer paid at professional rates and approved under the validation protocol; the avatar carries no performer's face.
Private by design
For banks and hospitals the whole system runs on their own premises; audio never leaves the building, and raw recordings are deleted once processed.
Lines we will not cross
No deception detection, no mass screening, no use the community formally objects to; that is written into how the data may be licensed.
"Nothing about us, without us." Written into the UN Convention on the Rights of Persons with Disabilities, and into how we work.
Wherever people need to understand each other.
The same avatar, the same library, the same conversation. Only the room changes.
Banks and public offices
Walk in, ask your question, get an answer. At the counter, without a booking.
Hospitals and clinics
Symptoms, consent, instructions. Understood, in private, at the moment they matter.
Airports and stations
Gate change, delay, platform. Announced in sign language on the screen, not just over the speakers.
Schools and universities
Lectures, parent meetings, the school play. Every child's family in the room.
Broadcast and streaming
News and shows with a signing avatar track, the way subtitles are today.
In your pocket
A phone, a stranger, a question. A conversation on the street that used to be impossible.
Why now
Accessibility is now a legal obligation for organizations that serve the public: the Americans with Disabilities Act and Section 504 in the United States, the European Accessibility Act across the EU since June 2025, and national equality and sign language acts. These laws ask for effective communication, which for many Deaf people means their own sign language. SignaNova makes that scalable for the routine interactions where human interpreters cannot match the volume, with human interpreters retained where the Deaf community asks for them.
Where we are.
Working application since August 2026
Speech and text to a signing avatar, American Sign Language.
Croatian Sign Language during 2027
Recorded with Deaf native signers and validated with the community; two pilot deployments in Croatia from April 2027.
Further European sign languages after
Same studio, same protocol, same runtime.
Patent pending
Four US provisional patent families.
See it for yourself.
Watch the preview, join a live session where someone on the call speaks unscripted and watches the avatar sign it back, or visit the studio in Zagreb.
Built in Croatia. Made for everyone.