nanosamur.ai
GitHub
Private Speech AI

Your voice.
Your control.

Guarding your sensitive conversations.

nanosamur.ai is a complete open-source speech AI platform for organisations that cannot send sensitive conversations to a third party. Capture, transcribe, refine, and process speech entirely inside infrastructure you control.

View on GitHub Get started

Supported models

  • Whisper
  • Qwen
  • Nemotron
  • Parakeet
Live & Batch
Transcription
Speaker diarization
Agentic workflows
Open source
Webhooks
Self-hosted

See it in action

Product tour10:49

From conversation to insight

Speaker-aware transcripts, workflows, playback, and a searchable final record. Follow a session from start to finish.

Download product tour

Built for sensitive workflows

Useful from the first word

Follow the conversation live, then improve the transcript automatically as more context becomes available.

Know who said what

Speaker-aware transcripts, word timings, and synchronized playback make conversations easier to review and verify.

Turn conversations into action

Send transcripts to your own AI workflows, applications, and webhooks during or after a session.

Keep a complete record

Store recordings, refined transcripts, and final session records for search, replay, and downstream processing.

A complete operational platform

Applications included

Use the browser interface or Windows desktop application to capture, review, and replay conversations.

Ready to integrate

Connect through REST, realtime WebSockets, the Python SDK, Kafka events, workflows, and webhooks.

Designed for real workloads

Distributed processing stages, multitenancy, and persistent records provide a foundation for shared services.

Observable end to end

Follow a session across the stack with correlated metrics, logs, and distributed traces.

Runs where your data has to stay

Private infrastructure

Run the complete platform on-premises, in a private cloud, or on Kubernetes using infrastructure operated by your organisation.

Isolated environments

Process speech without cloud APIs, including in networks with restricted or no external connectivity.

Built for conversations that cannot become someone else’s data

Government & defence

Transcribe interviews, briefings, meetings, and operational debriefs inside controlled infrastructure, including isolated networks.

Healthcare

Capture consultations and clinical discussions while keeping audio and transcripts inside infrastructure governed by your organisation.

The whole product is open source

Run it yourself

Clone the Apache-2.0 repository and start the complete platform locally with Docker Compose.

cp .env.example .env
docker compose pull
docker compose up -d

Inspect, adapt, and extend

Not just a model or SDK. The browser UI, desktop application, APIs, orchestration, speech services, and deployment configuration are available as open source.

Get started → Read the docs → View GitHub →

Sensitive conversations stay inside your environment

No external transcription dependency

Capture, transcription, storage, and downstream processing can all run inside infrastructure operated by your organisation.

Own the data path

Keep recordings, transcripts, application data, and operational telemetry under your organisation’s policies and controls.

Run nanosamur.ai yourself

Clone the repository and start the complete speech AI stack on your own hardware.

View on GitHub Get started