Live transcription while the call happens.

Read along during a call

Keep a text preview beside your meeting while you listen. The floating window updates as speech is recognized, with your microphone and the other side of the call shown separately. Use Copy if you want to grab the preview text during the conversation.

Live transcription is experimental and off by default. It is a useful way to follow along, but words can change as recognition catches up.

How to enable live transcription

1. Open Config → Models and find Live transcript preview (optional).
2. Choose and install one of the live speech models available for your Mac. If the selected model also needs Silero VAD, install it under Silence detection.
3. Open Config → Experiments and enable Show live transcript.
4. Start a recording. The live transcript appears in a floating window alongside your call.

The live model is a separate choice from the model used for your finished transcript. There are three:

• Parakeet EOU 120M (the default): English. It is light enough for any Mac, Intel included, and detects the end of each sentence on its own, so it needs nothing else.
• Parakeet Unified 0.6B Streaming: English with punctuation. Experimental, Apple silicon only, and needs Silero VAD.
• Nemotron 3.5 Multilingual 0.6B: 28 languages, Czech included, detected automatically. Experimental, Apple silicon only, and needs Silero VAD.

The technology page rates each model and lists the languages Nemotron covers.

A live preview and a saved transcript

The preview text is not saved automatically. After you stop recording, echo99 creates the full transcript from the recorded audio through its normal transcription process, including speaker labeling. That is the transcript you can revisit, search, export, and summarize.

Closing the live window stops the preview while the recording continues. To bring it back, click the Live transcript row in the echo99 menu-bar panel while you are still recording. If a live model is unavailable or the preview falls behind, audio recording can still continue independently.

Local processing, with practical limits

• The speech models need an initial download. Once installed, the live preview can run offline without sending your audio to a transcription service.
• Noise, overlapping voices, and unfamiliar names can produce mistakes.
• Recognition takes time and uses your Mac's resources. The preview may lag or miss words; use the finished transcript and recording to check anything important.

Setting up your first call? Start with the getting started guide.

Related guides: Local transcription, Meeting summaries