---
title: "Transcribe interview recordings"
description: "Turn interview recordings into Word transcripts, provide study context, check speaker labels and timestamps, and find the saved files."
---

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

# Transcribe interview recordings

Use InsightGig to turn interview recordings into Word transcripts saved in your study folder. Review the transcript against the recording before using quotes or starting analysis.

This walkthrough uses a fictional snack-shopping interview in Consumer study. The 74-second recording uses two synthetic voices and contains no real participant or company details.

The example finishes with `Recordings/snack-shopping-demo.docx`, including time ranges and Moderator/Respondent labels. Your recording, speakers, and processing time will differ.

## Before you begin

Have these ready:

- A connected study folder containing the recording you want to transcribe.
- A connected Nexus account with enough balance and available spending limits.
- A discussion guide, research brief, or short description of the interview, if available.

Follow [Add and switch projects](/projects-and-files/open-and-switch-projects/) for project setup. Follow [Sign in to the desktop app](/get-started/sign-in-to-desktop/) if your account needs connecting.

To follow the example, download [the snack interview recording and brief](/downloads/snack-interview-recording.zip). Right-click the download in File Explorer and select **Extract All**. Name the extracted folder `Consumer study` and check that it contains `Recordings` and `Research Instruments`.

Open `Consumer study` using [Add and switch projects](/projects-and-files/open-and-switch-projects/#open-the-folder-picker).

:::caution[Check whether you can send the recording]
Transcription sends recordings to an online service for processing. Use recordings and supporting context approved for that use.
:::

## Prepare the recording and request

1. In **Home**, select the project you added in [Before you begin](/research-workflows/transcribe-interviews/#before-you-begin). For the example, choose **Consumer study**. Select **New session**, then choose **Research** in the message box.

   Research can carry out a transcription request directly. Leave **Web** and **Desk Research** off for this task.

   ![Message box controls showing Research with Web and Desk Research switched off](@/assets/images/research-workflows/transcription-chat-controls.png)

2. Select **Open project files** and expand the folder containing the recording. The example uses `Recordings/snack-shopping-demo.wav` and `Research Instruments/transcription-demo-brief.md`.

   ![Recordings folder containing snack-shopping-demo.wav and Research Instruments containing the transcription context brief](@/assets/images/research-workflows/transcription-source-files.png)

3. In **Ask anything...**, name the recording and any supporting brief. Use this request for the example, then select **Send**.

   ```txt
   Transcribe Recordings/snack-shopping-demo.wav for the Consumer study.
   Use Research Instruments/transcription-demo-brief.md for context.
   First check the recording and tell me where the Word transcript will be saved.
   Wait for me before starting transcription.
   ```

   Ask the agent to check the recording and wait before starting. This gives you a checkpoint to confirm the selection and output location.

   ![Request naming the recording and context brief, with instructions to check the file and wait before transcription](@/assets/images/research-workflows/transcription-request.png)

A useful context brief describes the study objective, interview type, languages, specialist terms, and known speaker roles. Include factual background that helps interpret the recording.

For example, identify a moderator and one respondent, or explain a product demonstration discussed on screen. Do not invent identities or include unnecessary personal details.

If you have no supporting document, describe the interview in the conversation. The agent may ask for context; you can also ask it to proceed without context.

You should see the recording check and a proposed Word output location. The first check does not upload or transcribe the recording.

## Check the selection and start

1. In the agent's reply, confirm the recording count, duration, and output location.

   For the practice files, check that the reply names `snack-shopping-demo.wav`. The recording is 74 seconds long. Confirm the proposed Word filename and location. The example saves `snack-shopping-demo.docx` beside the recording.

   Check that the agent has understood the study context. Correct the file selection or background before continuing if either is wrong.

   ![Conversation with the recording check, the Word output location, and confirmation that the agent is waiting](@/assets/images/research-workflows/transcription-selection-check.png)

2. Reply in the same conversation when you are ready to start.

   In this example, the reply is `Go ahead and transcribe this recording using the context brief. Save the Word transcript beside the recording.`

   If InsightGig presents an approximate charge for approval, review it before agreeing. The estimate is not a guaranteed final charge or spending cap.

   ![Reply asking InsightGig to proceed using the context brief and save the Word transcript beside the recording](@/assets/images/research-workflows/transcription-start-reply.png)

3. Follow the **Transcription** activity and wait for completion.

   Expand the activity to inspect filenames and their status when the app shows a progress card. It can show preparation, uploading, transcription, and Word rendering.

   The screenshot shows the finished example: **Transcribed**, `Transcription completed`, and `Complete` beside the recording. Do not treat an estimate as a completed transcript.

   ![Expanded Transcribed activity showing Transcription completed and Complete beside snack-shopping-demo.wav](@/assets/images/research-workflows/transcription-completed-file.png)

Check the completion message for a link to the Word transcript. Use this message if you only see a **Transcription** entry in the conversation's activity list. The document is already saved on your computer.

The service checks your account balance and spending limits before starting. Some eligible recordings can proceed without separate charge approval unless you explicitly ask the agent to wait.

If your balance is low, follow [Buy credits](/console/usage-and-billing/fund-console-account/). If a spending limit blocks the run, follow [Manage model access and budgets](/console/workspace/manage-model-access/). Then ask the agent to check the recording again.

## Open and check the transcript

1. In the completion message, select **Open the Word transcript**, or the corresponding file link in your conversation.

   A preview tab opens in InsightGig. You can also find the saved `.docx` file in the folder view panel.

   ![Completion message confirming the transcript is saved locally and offering Open the Word transcript](@/assets/images/research-workflows/transcription-open-word.png)

2. Read the transcript in **Preview**.

   Each passage has a time range, a speaker label, and the transcribed words. Moderator passages are bold in this example.

   Use [Find and preview project files](/projects-and-files/folder-view-panel/) to find the saved file in Windows or open it in Word.

   ![Word transcript preview showing the interview's timestamped Moderator and Respondent passages](@/assets/images/research-workflows/transcription-word-preview.png)

3. Play sample passages from the original recording and compare them with the Word text.

   Check the opening, speaker changes, important figures, and any unclear speech. A timestamp shows how much time has passed since the recording began. It can help you find the passage in your audio player.

   In the practice recording, listen around 19 to 26 seconds for the respondent's price range of `20 to 30 rupees`. Your transcript may place the passage a little earlier or later. Confirm the amount and speaker against the recording.

   ![Transcript excerpt with the moderator question and the respondent's 20 to 30 rupee price range highlighted](@/assets/images/research-workflows/transcription-check-figures.png)

You should have a saved transcript that you have checked against its source. Keep the recording so you can verify later quotations and corrections.

Speaker labels and transcription can be wrong, especially with overlapping voices or unclear audio. A completed run confirms delivery, not research accuracy.

For more preview controls, follow [Find and preview project files](/projects-and-files/folder-view-panel/).

## Transcribe a folder of recordings

After checking one recording, name the specific folder when you want separate transcripts for a batch. The folder may contain nested folders.

For example, type `Transcribe the recordings in Recordings. Use the discussion guide for context. First check the selection and wait before starting.`

Check the file count before proceeding. Do not request the whole study folder if you only intend to process one interview group.

The output location depends on what you select:

| Selection | Word output |
| --- | --- |
| `Recordings/interview-01.wav` | `Recordings/interview-01.docx`, beside the source recording. |
| The `Recordings` folder | Separate Word files in the sibling `Recordings-transcripts` folder. |
| `Recordings/Group A/interview-01.wav`, within a folder selection | `Recordings-transcripts/Group A/interview-01.docx`, preserving the nested folder. |

Keep the original filenames meaningful so you can match each transcript to its recording. The transcription workflow does not overwrite an existing unrelated Word file.

## Choose processing for a video interview

If your selection contains video, the agent asks whether visual information matters. Explain what the recording shows before choosing a processing mode.

Compare the two choices:

| Choice | When it fits | What to consider |
| --- | --- | --- |
| Audio-only processing | A speech-led interview or call where the conversation provides the needed evidence. | Uses the audio. It cannot use visible name labels, slides, or demonstrations. |
| Video processing | A focus group, concept test, or demonstration where visible material helps explain speech or identify speakers. | Uses visual evidence as well as audio and usually costs more. |

For example, video can help when a respondent points to a pack or discusses a concept card without reading it aloud. Visual evidence still needs checking.

Choose for the recording's content. A video file does not automatically require video processing, and a video without an audio track cannot be transcribed.

## Stop a transcription run

While the agent is working, select **Stop** in the message box to cancel the active turn and its hosted transcription work. Temporary cloud files are cleaned up.

If no agent turn is active, use the command palette to find **Cancel transcription**. The command palette searches app actions, files, and conversations.

The default shortcut is `Ctrl+Shift+P` on Windows or `Cmd+Shift+P` on Mac. Search for `Cancel transcription`, then select that action.

The confirmation offers **Keep running** and **Cancel transcription**. Choose **Keep running** if you do not want to cancel.

![Cancel transcription confirmation explaining that it stops hosted work from the chat and removes temporary cloud files](@/assets/images/research-workflows/transcription-cancel-dialog.png)

Cancellation does not undo charges for successfully processed parts. Keep any completed Word transcripts and check which recordings still need processing.

## Resolve a recording problem

Use the reported error to decide the next step:

| What happens | What to do |
| --- | --- |
| No transcribable media is found | Check the selected file or folder and whether the recording can be opened. A video needs an audio track. |
| A file exceeds a limit | Select a recording within the three-hour and 3 GB per-file limits. A folder selection supports up to 1,000 recordings. |
| An existing Word document blocks output | Preserve that document, rename or move it, then ask the agent to check the recording again. |
| The selection changed after the check | Ask for another check before starting with the changed files. |
| Some recordings fail | Review completed outputs and the listed failures. Ask the agent to check the same selection again before retrying unfinished work. |
| The app closes or loses its connection | Reopen the same project and conversation. Ask the agent to continue transcription for the same recording or folder. |

Do not repeatedly submit the same request while a run is active. If a previous run exists, InsightGig can reconnect to it.

Do not convert a recording merely because of its filename extension. If InsightGig reports a genuine decoding problem, preserve the original before creating a compatible copy.

## Next steps

Use the reviewed Word transcripts as source material for [Analyze qualitative data](/research-workflows/analyze-qualitative-data/). Keep corrections and source references with the study.

Use [Check credits and usage](/console/usage-and-billing/track-usage/) to review recorded usage after the task.

Source: https://docs.insightgig.ai/research-workflows/transcribe-interviews/index.mdx
