Skip to content

Getting Started

github-actions[bot] edited this page Oct 24, 2026 · 1 revision

Getting started

The welcome guide

The first time Sabhalekh opens, it walks you through four short steps. You can skip it at any time and change everything later in Settings.

Welcome guide

  1. Welcome. A reminder that everything stays on your PC.

  2. Hardware. Sabhalekh looks at your PC and shows where speech recognition will run: your graphics card, your processor's built-in graphics, or the CPU. Green ticks mean the speech models are ready.

    Hardware check

  3. Audio test. Click Start test, then say something and play any sound (a YouTube video, music) for four seconds. Both lines should turn green:

    • Microphone works: Sabhalekh can hear you.
    • System audio works: Sabhalekh can hear what your PC plays, which is how it hears the other people in a call.

    If one stays red, see Troubleshooting.

  4. Test recording. Click Record 10 seconds and say a sentence in English or Hindi. A few seconds later your first transcript appears in the list on the left.

A tour of the window

The main window

On the left (the sidebar):

  • New Recording starts recording a meeting.
  • Upload Media transcribes audio or video files you already have.
  • Search all meetings looks through every transcript at once.
  • Filter and Date narrow down and sort the list.
  • Your meetings, newest first. Click one to open it.
  • At the bottom: the hardware badge (for example GPU Mode: GTX 1650), the processing queue, and Settings.

On the right: the open meeting.

  • The title at the top. Click it to rename the meeting.
  • The steps (Audio extracted → Transcribed → Diarized) and the talk time of each speaker.
  • The transcript: each line has the speaker's name, the time, the words and the language (EN or HI).
  • The player at the bottom: waveform, a coloured bar showing who spoke when, play/pause, skip ±10 seconds, speed and volume.

What happens after you record or import

  1. The meeting appears in the list straight away, marked as processing.
  2. Sabhalekh extracts the audio, finds where people speak, works out who is who, and writes down the words.
  3. When it's done you get a Windows notification, and the transcript opens.

A one-hour meeting takes about 5 minutes on a laptop with a graphics card and about 10 minutes on the CPU. You can keep working, and keep recording, while it processes.

Next steps

Clone this wiki locally