A Breakthrough in Accessibility: Real‑Time Speech‑to‑Sign in Africa

In 2025 a Kenyan startup named Signvrse (based in Nairobi) was awarded the prestigious Africa Prize for Engineering Innovation for its flagship app, Terp 360. The app uses AI, motion‑capture and 3D avatars to translate spoken or written language into Kenyan Sign Language (KSL) almost instantly.

Why this matters

Huge access gap: In Kenya alone, the deaf and hard‑of‑hearing community faces major barriers in education, healthcare, employment and public services due to lack of interpreters.
Technological leap: Terp 360 records sign language interpreters in motion‑capture suits, builds a growing dataset (more than 2,300 signs recorded so far), and generates lifelike avatars that “sign” the translated statement.
Scaling potential: Because human interpreters are scarce and expensive, this kind of tech could scale far more broadly — “interpreters at scale” rather than one person per service point.
Recognition and momentum: Winning the Africa Prize brings funding (the award is ~£50,000), visibility, mentorship and pathway to growth — helping push an African‑developed accessibility technology onto a global stage.

How it works (in simplified terms)

The user speaks or types a phrase in a supported language (initially English, possibly Swahili in future).
The system’s AI recognises the speech/text, maps it to a sign language equivalent using the dataset of gestures captured from real interpreters.
A 3D avatar (designed to look realistic, with correct hand orientation, face expression, body movement) renders the sign language output.
The output can be delivered via web, mobile screen, or integrated into a service (health‑care check‑in kiosks, education tools, public‑service websites).

What’s been emphasised in coverage

The founder, Elly Savatia, is young (born 2001) and has positioned the app as “Google Translate for sign language.”
Collaboration with deaf community in Kenya: the dataset is not built in a vacuum — real signers, real language variation are part of the build.
Vision to expand beyond English/KSL, to other African languages + sign languages across the continent by 2027.
The award recognises technical novelty, social impact potential and African‑origin engineering.

What the Story Doesn’t Fully Cover (but matters)

While the core narrative is impressive, here are some layers and caveats to consider:

1. Language & dialect complexity

Sign languages are not simply manual translations of spoken languages; they have unique grammar, spatial syntax and regional variations. Kenyan Sign Language (KSL) itself may vary by region.
Translating from speech/text into sign is only one direction. The reverse direction (signed‑to‑speech or signed‑to‑text) is far more complex and less mature.
The app currently supports English → KSL (and as some reports state possibly Swahili → KSL soon). Many users speak other languages/dialects (for example Sheng, local Kenyan English variants), which may lead to mismatches or misinterpretation.

2. Accuracy & usability

Early user‑testing indicates the app still has “errors” in sign‑avatar rendering, missing nuance (face expression, finger spelling, spatial context) and benefits hearing users more than deaf users (in sign‑to‑voice mode).
Real‑time translation in messy environments (poor audio, rapid speech, mixed ambient noise) remains more of a challenge—motion capture, AI training must handle variation.
The avatar sign output may not be ideal for all contexts: e.g., health‑care dialogues, legal interpretation, where accuracy and nuance matter.

3. Deployment, cost, infrastructure

Scaling the tech to mobile offline use, low‑bandwidth settings, and remote rural areas (without stable internet) is non‑trivial.
Integration into healthcare, education, government service systems requires training, institutional buy‑in, regulatory frameworks, and assurance of reliability.
Business model: While the app may have B2B (corporate, institutions) opportunities, the affordability for individual users (especially in low‑income settings) could be a barrier.

4. Sustainability & local capacity

Building the large dataset of sign language gestures is labor‑ and specialist‑intensive. The startup will need funding, talent, and ongoing updating to handle new vocabulary, slang, evolving sign usage.
The tech must stay culturally relevant (e.g., local sign variants, regional dialects, context‑appropriate gestures) and not rely purely on Western sign‑language templates.
Ensuring deaf communities are co‑designers, not just “users,” matters for accessibility ethics.

What This Means & What’s at Stake

For Deaf and Hard‑of‑Hearing Communities

Potential to dramatically improve access: imagine hospitals where the interpreter is built‑in, classrooms where deaf learners get live sign translation, public‑service announcements signed instantly.
It reduces reliance on human interpreters who are scarce and expensive — opening possibilities for more inclusive workplaces, education and civic participation.

For Technology, AI & African Innovation

This is a strong example of African‑origin innovation solving a locally relevant problem (and globally applicable too).
It positions Kenya (and Africa) in the global assistive‑tech ecosystem, not just as a “user” but a “creator.”
The dataset‑building work (motion capture, sign language translation) also contributes to broader AI research on “low‑resource languages” and multimodal translation.

For Institutions: Education, Healthcare, Government

Sign‑language translation tech opens policy windows: governments may consider requiring sign‑accessibility in public services, broadcast, web content, and may partner with tools like Terp 360.
Institutions will need to evaluate the reliability of AI sign‑translation, its suitability in high‑risk settings (legal, medical), and ensure deaf users are not left with lower‑quality “AI‑only” solutions vs human interpreters.
Scaling such tech could significantly reduce costs of providing sign‑language interpretation and improve service‑reach.

Frequently Asked Questions (FAQ)

Q: What exactly does the app translate?
It translates spoken or written input (initially English, possibly expandable to Swahili) into Kenyan Sign Language (KSL) using animated avatars, so the output is signed.

Q: Can it translate sign language to speech/text?
Not yet fully. The current focus is on speech/text → sign. The reverse direction (sign → speech or text) is much more challenging (vision/gesture recognition, context, grammar) and is still under development.

Q: How accurate is the translation?
It’s promising but not perfect. Users and experts report some errors, especially with complex phrases, fast speech, or when context‐specific signs are needed. The tech is still improving.

Q: Will this replace human sign interpreters?
Not entirely—and not immediately. For high‑stakes settings (court, legal, psychiatric, some medical situations) human interpreters remain critical because they bring nuance, context, cultural sensitivity, and judgement. The app is more likely to augment or expand access rather than fully replace all human interpreters.

Q: Which languages and sign languages are supported?
Currently the Kenyan app supports translation into Kenyan Sign Language (KSL). Input languages include English, with plans for Swahili. Expansion to other African sign languages is anticipated by around 2027.

Q: How can institutions adopt this kind of technology?
Education systems, hospitals, and corporate workplaces can integrate the platform (e.g., via API, mobile/web version). But they will need to validate it, train staff, ensure user experience (deaf users review), and consider infrastructure (internet, mobile devices).

Q: What are the main barriers to widespread use?
Key barriers include: cost of mobile/internet access for users, quality issues in translation for complex contexts, need for dataset expansion (new signs, slang, dialects), ensuring offline capability in low‑connectivity areas, and ensuring the technology is culturally inclusive.

Final Thought

The development of Terp 360 and its recognition with a top innovation prize is more than a cool tech story. It signals a shift: access to communication is a human right, and AI can help bridge deep divides — but only if designed with the communities in focus, scaled responsibly, and deployed ethically.

For Africa’s deaf and hard‑of‑hearing communities, this may mark a turning point toward more inclusion. For global tech and assistive‑innovation ecosystems, it shows that high‑impact, culturally rooted solutions are emerging from the continent. The journey ahead — in accuracy, scale, accessibility and trust — remains. But the sign is clear: communication barriers are being challenged, one avatar at a time.

Sources CNN