Proto continues to strengthen Kinyarwanda voice support across its AI agent deployments in Rwanda, improving access to critical services in government and financial sectors where local language communication is essential for trust and service delivery.
Supported by ScaleAI in Canada and the Gates Foundation, Proto’s Kinyarwanda voice capability combines automated speech recognition (ASR) and text–to–speech (TTS) to enable natural, accurate interactions for Kinyarwanda speakers.
Kinyarwanda is widely spoken, yet commercial voice automation has historically lagged because low–resourced languages face a familiar barrier – usable training data is scarce, everyday speech varies by context, and real service conversations often include mixed terminology. These constraints are especially visible in high–stakes moments like scam reporting and complaint redress, where clarity and speed matter.
Deployment by Rwandan institutions
Kinyarwanda voice capability is already in use within national–level systems in Rwanda.
In the financial sector, it supports complaint processing and consumer guidance for the National Bank of Rwanda, helping standardise how issues are captured, routed, and resolved across the market.
In public services, it is being applied through Mbaza – the citizen engagement initiative with the Ministry of Local Government (MINALOC) – designed to extend consistent support across Rwanda’s 30 districts.
While these national deployments show the public value at scale, the same feature is not limited to government programmes – it is available to any organisation that signs up for Proto and enables AI voice agents in its channels.
A real production sample generated by Proto’s Kinyarwanda text-to-speech model. The message says "You can visit agricultural research centres or the local authorities responsible for agriculture in your district to receive guidance and advice. You may also attend training sessions or join farmers’ cooperatives to gain experience and support."
A real production sample generated by Proto’s Kinyarwanda text-to-speech model. The message says "In Rwanda, you can engage in tourism at natural attractions such as Volcanoes National Park – home to mountain gorillas – Akagera National Park, Nyungwe National Park, and Lake Kivu. To join the Visit Rwanda programme, you can contact the Rwanda Development Board – they will assist you in organising and registering your activities."
Local language AI from the ground up
There was no off–the–shelf Kinyarwanda model that could meet the reliability and audit requirements needed for institutional deployment. Proto built the Kinyarwanda voice layer essentially from scratch and fine–tuned it using a dedicated Kinyarwanda dataset delivered through Voices.com as part of the ScaleAI consortium support.
The work addresses structural gaps common to low-resource languages, including scarce labelled data, inconsistent orthography in written Oshiwambo, and the absence of established evaluation benchmarks. Rather than adapting models trained on high-resource languages, the Oshiwambo voice layer was developed and validated specifically for these conditions by Proto’s R&D team led by CTO Weiying Kok.
Benchmark snapshot
Text-to-speech – Kinyarwanda
Speech-to-text – Kinyarwanda
These performance metrics reflect the scale of achievement. Accuracy and fallback rates are improving continuously as additional real-world Kinyarwanda voice data is captured through active deployments and incorporated into supervised evaluation and retraining cycles.
Built for messaging–led service delivery
Kinyarwanda voice plays a practical role in the digital transformation underway in emerging markets like Rwanda, where messaging platforms are increasingly central to communication and commerce. Organisations can enable voice in the channels people already use -such as WhatsaApp, Messenger - and keep workflows consistent across service teams.
Interactions adapt automatically – if someone sends a voice message, the AI agent replies with audio; if they switch to text, the replies switch to text too.
Beyond consumer protection
Beyond financial consumer protection, Kinyarwanda voice capability can support high–impact sectors such as agriculture, healthcare, and education. It can help deliver timely market prices and weather updates, guide people through services remotely, and support students in rural areas with accessible learning information.
Proto’s work with the NBR reflects a commitment to underserved populations, digital inclusion, and financial literacy – using AI voice agents to help institutions handle cases more efficiently, reduce delays, and extend access beyond urban centres.
The Kinyarwanda rollout forms part of Proto’s broader commitment to building inclusive AI infrastructure, ensuring that underserved populations are excluded from the world’s newest economic opportunities.
