Unamentis
Fiscal Host: Open Source Europe
UnaMentis is a Open Source, mobile ,voice, AI learning tool that allows full extended hands free voice learning sessions with existing curriculum.
Contribute
Become a financial contributor.
Financial Contributions
Recurring contribution
Backer
Covers roughly half the monthly storage, monitoring, and data transfer costs for the beta hosting environment. You're directly funding the infrastr... Read more
€25 EUR / month
Projects
Support the following initiatives from Unamentis.
Project
Presenting UnaMentis at open source conferences, education technology events, and AI meetups is h...
Project
separate hosting environment for testing new features, provider integrations, and performance op...
Project
We need 2 fairly new Android phones for hands on testing of our Android client in a mobile setting.
Project
Pay for cloud hosting to enable our initial small group public beta testing.
Top financial contributors
Richard Amerman
€100 EUR since Dec 2025
Unamentis is all of us
Our contributors 2
Thank you for supporting Unamentis.
Budget
Transparent and open finances.
-$43.08 USD
~€36.82 EUR
Pending
operations
-$8.34 USD
~€7.13 EUR
Pending
operations
-$12.99 USD
~€11.10 EUR
Pending
infrastructure
€
Today’s balance€87.44 EUR
Total raised
€87.44 EUR
Total disbursed
--.-- EUR
Estimated annual budget
€300.00 EUR
Connect
Let’s get the ball rolling!
News from Unamentis
Updates on our activities and progress.
Launching the Beta: Where We Are and What's Next
UnaMentis is about to hit a major milestone. The iOS beta launches in the next week, and we wanted to share where the project stands and what's ahead. What's built and ready · The iOS app is complete and ready for beta test...
Published on April 8, 2026 by Richard Amerman
Keeping the momentum up
The UnaMentis github repo(s) are now in their own org. Works has started on an Android client. IOS is still our MVP, but it is important to get the Android client on track. We are also working on a web client. While we think the best perfor...
Published on January 6, 2026 by Richard Amerman
Expanding Colaboration
We are first and formost an Open Source project intent on helping individuals expand their knowledge and capabilities. While we finish our MVP work, which is only expected to take another week or two, we are limiting direct participation/co...
Published on December 29, 2025 by Richard Amerman
About
UnaMentis is an open source, voice-first AI learning platform. It delivers structured educational content through natural, fully interactive voice conversation, enabling hands-free learning while walking, driving, or doing chores. The platform is MIT-licensed, provider-agnostic, and designed to run entirely on-device when needed, with zero dependency on any single cloud provider.
The name means "one mind" in Latin.
The core experience is full bidirectional voice interaction with barge-in capability. Learners can interrupt at any point to ask questions, explore tangents, challenge an idea, or request examples, then return seamlessly to exactly where they left off in the curriculum. The AI adapts to the learner in real time. This is not an audiobook or a chatbot. The learner is in control of the experience at all times.
How this started
In April 2025, the project's founder discovered the transformative potential of AI-assisted learning through voice conversation. Detailed, interactive learning sessions, where you could pause, ask questions, explore tangents, and then resume exactly where you left off, proved to be an extraordinarily effective way to learn. But the experience was inconsistent across platforms, dependent on a single provider's decisions, and could break overnight with a product update. There was no open alternative. So we decided to build one.
What exists today
UnaMentis is working software, not a concept.
iOS App (in beta). A fully functional voice learning client with real-time voice interaction, curriculum browsing, and session management. Android and web clients are in development.
Provider-agnostic architecture. 9 speech-to-text providers, 8 text-to-speech providers, 5 LLM providers. Cloud, on-device, and self-hosted options. Any component can be swapped without code changes.
On-device voice synthesis. We ported Kyutai's Pocket TTS model to iOS using Rust and Candle, achieving near-server-quality voice synthesis running entirely on the user's device. No API keys required, no data leaving the phone, no network latency.
Open curriculum standard. The UnaMentis Curriculum Format (UMCF) is a JSON-based specification designed for conversational AI learning. Every text field has optional spoken variants optimized for TTS pronunciation. It maps to IEEE LOM, LRMI, SCORM, xAPI, QTI, CASE, and Open Badges.
Management server. Python-based server with WebSocket support for remote logging, real-time metrics, curriculum management, system monitoring, and hosted inference for connected clients. React-based operations console with Curriculum Studio and Voice Lab.
Voice-first with image support. Sessions can include maps, diagrams, and illustrations within transcripts, combining voice learning with visual context.
Plans include expanding across platforms, supporting multiple languages, and integrating diverse curricula including MIT and K-12 sources. Users will be able to create custom learning materials or upload existing resources converted into structured curricula.
iOS App (in beta). A fully functional voice learning client with real-time voice interaction, curriculum browsing, and session management. Android and web clients are in development.
Provider-agnostic architecture. 9 speech-to-text providers, 8 text-to-speech providers, 5 LLM providers. Cloud, on-device, and self-hosted options. Any component can be swapped without code changes.
On-device voice synthesis. We ported Kyutai's Pocket TTS model to iOS using Rust and Candle, achieving near-server-quality voice synthesis running entirely on the user's device. No API keys required, no data leaving the phone, no network latency.
Open curriculum standard. The UnaMentis Curriculum Format (UMCF) is a JSON-based specification designed for conversational AI learning. Every text field has optional spoken variants optimized for TTS pronunciation. It maps to IEEE LOM, LRMI, SCORM, xAPI, QTI, CASE, and Open Badges.
Management server. Python-based server with WebSocket support for remote logging, real-time metrics, curriculum management, system monitoring, and hosted inference for connected clients. React-based operations console with Curriculum Studio and Voice Lab.
Voice-first with image support. Sessions can include maps, diagrams, and illustrations within transcripts, combining voice learning with visual context.
Plans include expanding across platforms, supporting multiple languages, and integrating diverse curricula including MIT and K-12 sources. Users will be able to create custom learning materials or upload existing resources converted into structured curricula.
Project stewardship
Our commitment to keeping UnaMentis free and open source is not just a policy. It is something we intend to protect permanently.
Our long-term goal is to establish a dedicated nonprofit foundation in Europe to permanently protect the project's name, intellectual property, and open source status, ensuring the project can never be sold, closed, or made proprietary. Until that structure is in place, we have partnered with Open Collective Europe, a Belgian nonprofit that provides fiscal hosting for open source projects. This allows us to receive donations transparently while maintaining our independence.
While hosted by Open Collective Europe, these commitments hold:
- The project is non-commercial and will remain free and open source
- No services, subscriptions, licenses, or paid access are offered
- Donations are used only for open-source development, infrastructure, and testing
What donations fund
All donations to this collective support the open source project directly:
- Hosting infrastructure: GPU-accelerated servers for LLM and TTS inference, so beta users can experience the full platform without needing their own API keys. This is the largest single cost, running approximately €370-€600/month depending on the number of users.
- Development and staging: A separate environment for testing new features and provider integrations before they reach production (~€130/month using spot instances).
- Testing across devices: Physical Android devices for real-device voice testing. Voice interaction timing, audio latency, and on-device inference behave fundamentally differently on physical hardware than in emulators.
- Operational costs: Domain registration, monitoring, support services, and development tooling (~€40/month).
- API testing: Validating integrations across multiple STT, TTS, and LLM providers to maintain the provider-agnostic architecture.
Cost controls are built into the platform: per-user and global usage limits, real-time cost tracking, and feature flags for rapid cost management. We run self-hosted models (not per-token cloud APIs) to keep costs predictable and low.
Transparency
Every donation and expense is tracked openly on this page. We believe that if you support this project, you deserve to see exactly where your contribution goes. All expenses are submitted through Open Collective with descriptions and receipts.
Our team
Richard Amerman
Admin
Cy Goerdt
Admin