We preserve the voices of indigenous communities worldwide through open-source frugal AI voice models — built by volunteers, researchers, and community champions.
When a language disappears, it takes with it an entire worldview — unique knowledge of nature, history, medicine, and culture that exists nowhere else. Today, over 40% of the world's languages are endangered, many spoken by only a handful of elders.
The Saving Voices Project is a volunteer-led global initiative that harnesses the power of frugal AI — lightweight, accessible voice models — to document and preserve these languages in a form communities themselves can use and control.
Local champions and volunteers gather authentic voice data directly from community members and elders.
Researchers build lightweight, open-source speech models that run on low-resource devices — accessible to all.
All data and models are freely shared back to communities and the global research community.
A language is not just words — it is the living memory of a people, encoded in sound. When the last speaker falls silent, we lose something irreplaceable.— The Saving Voices Project
The Soliga are an indigenous community living in the forests of Karnataka, India. Our pioneering research documents their endangered language using frugal AI methods — a model for global indigenous language preservation.
One of Karnataka's oldest indigenous forest-dwelling communities, the Soliga have inhabited the Biligiri Rangaswamy Temple Wildlife Sanctuary for generations. Their language carries unique ecological and spiritual knowledge that cannot be translated.
Our team developed a frugal voice data collection pipeline working directly with Soliga community members. We recorded natural speech, oral traditions, and daily language use, producing a richly annotated dataset for AI model training.
Using the collected data, we built lightweight ASR and TTS models specifically designed to run on low-powered devices — ensuring the technology is accessible to community members regardless of infrastructure constraints.
The entire project was designed with and for the Soliga community. Community champions were trained as voice data collectors, ensuring ownership, consent, and cultural sensitivity throughout the research process.
The Soliga voice dataset is being made freely available to researchers, educators, and community members globally — a contribution to both AI research and indigenous language revitalisation efforts worldwide.
Our Soliga methodology is designed to be replicated across any indigenous language community globally. The tools, workflows, and models we developed provide a scalable blueprint for language preservation at scale.
A frugal AI approach to indigenous language documentation — building voice models for low-resource communities
The Saving Voices Project was founded by researchers at the intersection of AI, linguistics, and community development — united by a belief that technology should serve the world's most vulnerable voices.
CEO of Flipped.ai and Fellow of the Cambridge Judge Business School. CTO of the Frugal AI Hub at Cambridge Judge Business School, pioneering accessible AI for underserved communities.
Assistant Professor at IIIT Dharwad, specialising in indigenous community engagement and linguistic documentation. Leads community partnerships and field research across South Asia.
We partner with indigenous communities and appoint local community champions who understand the cultural context and speak the language.
Community champions collect high-quality voice recordings using simple tools — stories, songs, conversations, and everyday speech — with full community consent.
Researchers and volunteers use the data to train lightweight, open-source speech recognition and synthesis models designed for low-resource environments.
All datasets and models are published openly and given back to the community — enabling education, cultural preservation, and future research forever.
We're building a global community of volunteers, researchers, engineers, and community advocates. Whether you're a linguist, an AI researcher, a developer, or simply someone who cares — there's a role for you in this mission.