Ecoute is a live transcription tool for Windows that listens to both sides of a conversation at once. It captures your microphone and your speakers’ output simultaneously and produces a real-time transcript, labeling which channel each line came from. The project has about 6,000 stars on GitHub, is MIT licensed, and was still being updated in 2026.
Most transcription apps only hear your microphone, so half of every call is missing. Ecoute records the loopback of your system audio alongside your voice and runs both streams through OpenAI’s Whisper speech-recognition model locally, displaying a rolling transcript in a simple window with “You” and “Speaker” channels. That makes it handy for meetings, lectures, interviews and accessibility: you get a searchable text record of everything said on both ends, generated on your own machine rather than in a cloud service.
Because both audio channels are captured, Ecoute transcripts read like proper meeting minutes rather than one-sided notes, and the live display doubles as a captioning aid during the call itself. Recordings and transcripts stay on the local machine, since Whisper inference runs on your own hardware — a meaningful difference from cloud meeting assistants that route audio through third-party servers. Performance depends on the Whisper model size you choose and on whether a GPU is available, with smaller models trading a little accuracy for real-time speed on ordinary laptops.
Free. Ecoute is open source under the MIT license and performs transcription locally with Whisper, so there are no per-minute fees.
Windows users who want a free, private, real-time transcript of calls and meetings — including the other participants’ side of the conversation.