Free AI softphone that records, transcribes and prompts
Most of my working day happens on calls: customer calls through the office PBX, and meetings in Zoom, Teams and Meet. The tools for those two worlds never talk to each other. Softphones just dial. Meeting bots join your calls, upload the audio to someone else's cloud and charge per seat. So we built AI-Softphone : one desktop app for Windows, macOS and Linux that makes SIP calls, records any…
The AI-Softphone is a desktop application available for Windows, macOS, and Linux that enables SIP calls, records meetings, and transcribes conversations into searchable transcripts and summaries. It operates entirely locally and is free and GPL-licensed. Upon initiating a call, the app provides a real-time prompter that generates subtitles, translations, and suggested responses based on user-provided scripts or interview guides.
Pre-installed assistants cover subtitle generation, translation, objection handling, and interview assistance, all of which are customizable. Before a call, users can practice using the built-in practice mode that allows playback of previous recordings for optimizing the prompter. After the call, AI-Softphone generates a speaker-synced transcript with full-text search capabilities across all conversations.
Users can export transcripts as text or subtitles, receive summaries, extract action items, identify key figures, categorize content, flag potential issues, and receive scores based on user-defined criteria. The app also offers support for asking questions about past conversations. To utilize AI-Softphone with local models, users can install Ollama and select a model such as llama3.1:8b.
The app can then be configured to connect to the local Ollama (or vLLM) endpoint. If users prefer cloud-based services like OpenAI, Claude, Gemini, Mistral, DeepSeek, Groq, Deepgram, or AssemblyAI, they can input their own API keys, and the app will display costs for each call while allowing capped processing to prevent overspending.
The development of AI-Softphone is based on C++20, Qt 6/QML, and PJSIP, with a headless command-line interface for further customization. Developers and administrators can integrate AI-Softphone with their own infrastructure, as it supports tel: links for click-to-dial from browsers or CRM systems, a localhost-only REST API for managing calls, contacts, history, and accounts, and webhooks for call initiation, state changes, and call termination.
Notably, AI-Softphone emphasizes data ownership, with all data stored in user-controlled files, credentials stored in the operating system's keychain, and no external interfaces for data listening. The app is modular, allowing users to selectively enable or disable recording, transcription, write-ups, meeting capture, import features, and integrations, resulting in a basic softphone functionality when all options are disabled.
Currently, macOS requires Apple silicon for compatibility, and the app does not yet support team features such as central administration, shared storage, or reporting, with these functionalities slated for future development in an Enterprise edition. Users can download AI-Softphone from https://ai-softphone.com, review comprehensive documentation at https://docs.ai-softphone.com, and provide feedback, particularly from those operating their own private branch exchanges (PBX) or utilizing existing meeting-notes tools. The developers welcome user feedback, as a GitHub star would be greatly appreciated.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.