> ## Documentation Index
> Fetch the complete documentation index at: https://docs.wittify.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Live Transcribe

> Real-time browser transcription. Speak, watch the transcript appear, copy it, and save.

<Note>
  The **Live Transcribe** modal opens from the dashboard's **Live Transcription** card, the **Record Audio** card, or the topbar's **+ Transcribe Audio** dropdown (entries **Live Transcription** and **Record Audio**). Your browser handles both the live transcript and the audio recording, no setup needed.
</Note>

## What this modal looks like

A 700 px popup (850 px on `lg+` screens) split into two panels.

| Panel | What you see                                                                                                                  |
| ----- | ----------------------------------------------------------------------------------------------------------------------------- |
| Left  | The language selector, the audio player (appears after you stop), and the **Start recording** / **Stop transcribing** button. |
| Right | The live transcript area as you speak, plus a **Copy** icon.                                                                  |

The header carries the title **Live Transcribe**. A small **X** in the upper end-edge corner closes the modal.

## Using the modal

<Steps>
  <Step title="Open the modal">
    Click **Live Transcription** or **Record Audio** on the dashboard's **Start Transcribing** grid, or open **+ Transcribe Audio** in the topbar and pick one of those entries.
  </Step>

  <Step title="Pick a language">
    Choose the language you'll be speaking in from the dropdown. Switching the language mid-session restarts recognition in the new language.
  </Step>

  <Step title="Start recording">
    Click the purple **Start recording** button. Your browser asks for microphone permission the first time. If you deny it, the modal shows the toast **Microphone access denied or unavailable.**

    Two things start in parallel: the audio is recorded so you can play it back, and the engine streams final transcript chunks into the right panel as you speak.
  </Step>

  <Step title="Speak">
    The right panel placeholder reads **Start speaking to see transcription...** As soon as the engine is confident a phrase is final, it gets appended to the transcript with a space.
  </Step>

  <Step title="Copy as you go (optional)">
    Click the small **Copy** icon at the corner of the transcript area to put the current text on your clipboard. The icon flips to a green check for 2 seconds. If clipboard access fails, the modal toasts **Failed to copy to clipboard.**
  </Step>

  <Step title="Stop transcribing">
    Click **Stop transcribing**. Both the recognition stream and the audio recording stop. A small audio player appears on the start-edge panel so you can review what was captured.
  </Step>

  <Step title="Close to save">
    Closing the modal saves the session whenever there's transcript text:

    * A new file is created in your library with the name **Live Transcription HH:MM** (your local start time).
    * The audio duration is computed from when you started to when you stopped, formatted `M:SS`.
    * You're routed to the [Transcript Editor](/en/Systems/Speech-to-Text/Transcript-Editor) for that file.

    If the transcript is empty, the modal closes without saving.
  </Step>
</Steps>

## Button states

| What's happening        | Button label          |
| ----------------------- | --------------------- |
| Idle (before you click) | **Start recording**   |
| Recording               | **Stop transcribing** |

## Browser support

Live transcription depends on a feature your browser provides. If your browser doesn't support it, you can still record the audio and have it transcribed afterwards by uploading the saved recording.

| Browser                                   | Live transcript | Audio recording |
| ----------------------------------------- | --------------- | --------------- |
| Chrome, Edge, Brave (desktop and Android) | Yes             | Yes             |
| Safari (desktop and iOS)                  | Yes             | Yes             |
| Firefox                                   | No              | Yes             |

## Languages

Same 16 languages as [Upload Audio](/en/Systems/Speech-to-Text/Upload-Audio): English, Arabic, French, Spanish, German, Turkish, Urdu, Hindi, Chinese, Japanese, Korean, Portuguese, Russian, Dutch, Italian, plus Auto Detect (which falls back to English for live mode). Each language is sent to the engine with its regional variant (Arabic uses Saudi, English uses US, French uses France, and so on).

<Tip>
  Live recognition shows you only **final** chunks once the engine is confident, not partial guesses that change as you speak. If you want a polished pass with partials and speaker diarization, upload the saved recording through [Upload Audio](/en/Systems/Speech-to-Text/Upload-Audio) afterwards.
</Tip>

## What ends up in your library

When you close the modal with a non-empty transcript, the saved file shows up in **Your Files** with:

| Field          | Value                                                |
| -------------- | ---------------------------------------------------- |
| Name           | **Live Transcription HH:MM** (your local start time) |
| Source         | Live recording                                       |
| Duration       | Computed from start to stop, formatted `M:SS`        |
| Language       | Whatever you picked in the dropdown                  |
| Speaker labels | Off (live mode doesn't separate speakers)            |
| Subtitles      | Off                                                  |
| Status         | **Completed** (no background processing)             |

## Common questions

<AccordionGroup>
  <Accordion title="Live Transcription doesn't work in my browser.">
    You're probably on Firefox, which doesn't expose the live transcription feature today. Use Chrome, Edge, Brave, or Safari for live mode. Or use [Upload Audio](/en/Systems/Speech-to-Text/Upload-Audio) instead, that works in every browser.
  </Accordion>

  <Accordion title="Microphone access denied or unavailable. How do I fix it?">
    Click the lock icon next to the URL in your browser, find the microphone permission, and switch it to **Allow**. Reload the page and try again. On Chrome and Edge you can also visit `chrome://settings/content/microphone` to manage permissions per site.
  </Accordion>

  <Accordion title="The transcript doesn't show partials, only finished sentences.">
    That's intentional. Live mode only shows phrases the engine is confident about, so you don't see text changing as you speak. For partials, record first and then upload through [Upload Audio](/en/Systems/Speech-to-Text/Upload-Audio).
  </Accordion>

  <Accordion title="Can Live Transcribe handle two speakers?">
    Live mode treats your audio as a single channel and doesn't separate speakers. If you have multiple speakers, record the call and run it through [Upload Audio](/en/Systems/Speech-to-Text/Upload-Audio) with **Detect Speakers Automatically** turned on.
  </Accordion>

  <Accordion title="The accuracy in Live mode feels lower than the upload flow.">
    Live mode uses what your browser provides, which varies in quality from browser to browser. The upload flow uses the Faheem engine and is the right choice for accuracy, especially for Arabic and dialects.
  </Accordion>

  <Accordion title="I forgot to copy my transcript and closed the modal.">
    You didn't lose it. Closing the modal with non-empty text saves the file automatically. Look in **Your Files** in the sidebar for **Live Transcription HH:MM**, click it to open the [Transcript Editor](/en/Systems/Speech-to-Text/Transcript-Editor) where you can copy or export.
  </Accordion>

  <Accordion title="My audio is still in the modal but the transcript area is empty.">
    The transcription engine couldn't pick out any speech (too quiet, too noisy, or a language mismatch). Open the audio playback to verify your mic was recording, then try again, or upload the saved recording via [Upload Audio](/en/Systems/Speech-to-Text/Upload-Audio) which handles harder cases.
  </Accordion>

  <Accordion title="Can I switch language while recording?">
    Yes. Pick a different language from the dropdown and recognition restarts in the new one immediately. The audio recording continues uninterrupted, so the audio file you save covers the whole session.

    <Tip>
      For production-grade Arabic-first transcription with the lowest latency, the Faheem voice engine handles real-time streaming. That's what powers the [Voice Conversations](/en/Systems/AI-Agents/Manage-Agent/Voice-Conversations) live captions and is rolling out to the Live Transcribe modal in a future release.
    </Tip>
  </Accordion>
</AccordionGroup>