speakr: Transcribe Audio and Turn It Into Searchable Notes

Overview

Speakr is a self-hosted application for turning recordings into transcripts and organized notes. It combines audio capture and uploads with transcription, summaries, transcript search, and tools for asking questions about one recording or a library of recordings.

It is suited to people and groups who want control over where their recordings are stored and how they are shared. Transcription and language-model services can be configured, including self-hosted options, so deployment choices affect how much processing stays on your own infrastructure.

Key Features

• Record microphone and system audio, or upload audio files; in-app recordings stream to the server during capture.
• Connect to a range of transcription services, including self-hosted WhisperX and hosted providers.
• View speaker-labeled transcripts with playback synchronized to transcript timestamps.
• Generate configurable summaries and extract events or action items.
• Search and ask questions across recordings, with citations linked to transcript moments.
• Organize recordings with folders, tags, retention policies, and automated exports.
• Share recordings with other users and groups, with configurable permissions and public links.
• Use the REST API and signed webhooks to connect external tools and workflows.

Use Cases

• A family or club can keep recordings and transcripts in a shared library, with access managed through groups and tags.
• Teams can document meetings, search past discussions, and follow up on decisions and action items.
• Researchers and interviewers can preserve recordings, locate relevant transcript passages, and export notes to another system.
• Individuals can capture lectures, consultations, or personal notes and use custom prompts to format the resulting summaries.

Project Facts

• Language: Python
• License: AGPL-3.0
• Stars: 4.1k
• Forks: 343
• Topics: none listed
• Archived: no

Getting Started

The README recommends Docker Compose:

mkdir speakr && cd speakr
wget https://raw.githubusercontent.com/murtaza-nasir/speakr/master/config/docker-compose.example.yml -O docker-compose.yml
wget https://raw.githubusercontent.com/murtaza-nasir/speakr/master/config/env.transcription.example -O .env
# Configure the required service credentials in .env, then run:
docker compose up -d

Open http://localhost:8899. See the README and installation guide for configuration and provider-specific setup.

Alternatives

• meetily: Meetily focuses on recording and summarizing meetings from a desktop app, while Speakr organizes and transcribes audio recordings more broadly.
• vexa: Vexa uses bots to join video meetings and provides speaker-attributed transcripts through an API, rather than managing a general library of audio recordings.

Considerations

• Speakr is self-hosted, but using hosted transcription or language-model providers sends relevant data to those services. Review each provider's data handling if recordings are sensitive.
• Setup requires configuring transcription and text-generation services or running compatible self-hosted services. WhisperX voice profiles require the WhisperX backend and its GPU-enabled service.
• The README describes a lightweight Docker image that skips PyTorch; semantic search in Inquire Mode then falls back to basic text search.
• The project is licensed under AGPL-3.0 and also offers a commercial license. Review the license terms before incorporating it into a service or distributing modifications.
• The README labels agentic Inquire Mode as an opt-in beta, so treat that feature as experimental.