The TASK_TRANSCRIBE function transcribes audio or video files from Google Drive using the Deepgram API. This powerful function automatically converts spoken content to text and can either place the transcription directly in your spreadsheet or create a Google Doc with the transcribed content.
=TASK_TRANSCRIBE(header, file_link_cell, [output_doc_name], [celladdress])
Parameter | Data Type | Required | Description |
---|---|---|---|
header | string | Yes | Column header for the task template. This appears in the first row of the output. |
file_link_cell | string | Yes | The cell containing the link to the file to transcribe (e.g., "A1"). You should copy (ctrl+c) the file from Google Drive and paste in the referenced cell. |
output_doc_name | string | No | Optional. The name of the output Google Doc to create with the transcription result (e.g., "Transcription Result"). |
celladdress | string | No | Optional. The cell address to place the result (e.g., "C1"). If not specified, the cell to the right of the function cell is used. |
When you specify an output_doc_name
parameter:
=TASK_TRANSCRIBE("Meeting Transcription", "A1")
=TASK_TRANSCRIBE("Interview Transcript", "B5", , "D5")
=TASK_TRANSCRIBE("Lecture Notes", "C3", "Physics Lecture Oct 2024")
=TASK_TRANSCRIBE("Podcast Episode", "E7", "Podcast Transcript EP45", "G7")
Cause: The function output was modified manually or corrupted.
Solution: Recreate the function with =TASK_TRANSCRIBE().
Cause: The file_link_cell parameter is empty or the cell doesn't contain a valid link.
Solution: Ensure you've specified a valid cell reference and that the cell contains a Google Drive link to an audio or video file.
Cause: The JSON has been modified and no longer contains "TASK_TRANSCRIBE" as the function name.
Solution: Recreate the function with =TASK_TRANSCRIBE().
Cause: The specified celladdress doesn't exist or is invalid.
Solution: Check the cell address format and ensure it exists in the spreadsheet.
Cause: The Deepgram API key is not configured or is invalid.
Solution: Configure your Deepgram API key in the Sheet Workflows sidebar under Settings.
Cause: The audio file couldn't be processed, possibly due to format issues, corrupt file, or network problems.
Solution: Check that the file is a supported audio/video format, ensure it's not corrupted, and verify your internet connection.
Cause: The system couldn't access the file in Google Drive.
Solution: Ensure the file has proper sharing permissions and is accessible to your Google account.
Cause: There was an issue creating the Google Doc with the transcription.
Solution: Check your Drive permissions and available storage space, or try using a different document name.
Open the Sheet Workflows sidebar, go to Settings, select "Deepgram" from the providers list, and enter your API key. You can get a Deepgram API key by signing up at https://deepgram.com .
Deepgram can handle files up to several hours long, but for very large files, processing may take longer. For optimal performance, consider breaking very long recordings into smaller segments.
Deepgram's whisper-large model supports multiple languages with automatic language detection. Best results are typically achieved with English, Spanish, French, German, Italian, Portuguese, Dutch, Russian, Japanese, Chinese, Korean, and Arabic, among others.
If you don't specify an output_doc_name, the transcription is placed directly in the output cell in your spreadsheet. If you specify an output_doc_name, a Google Doc is created in the "Sheetworkflows/Docs" folder in your Google Drive, and a hyperlink to that document appears in the output cell.
No, the file must be accessible via Google Drive. If your file is stored elsewhere, you'll need to upload it to Google Drive first, then paste the link in the referenced cell.
After entering the TASK_TRANSCRIBE function, select the cell and click the "Run" button in the Sheet Workflows sidebar. The function itself only generates the task definition JSON.
Yes, TASK_TRANSCRIBE works with both audio and video files. For video files, it extracts and transcribes the audio track.
Transcription accuracy depends on audio quality, clarity of speech, accents, and background noise. The whisper-large model used by Deepgram provides strong results for clear audio but may have difficulty with poor quality recordings or heavy accents.
Yes, you can create multiple TASK_TRANSCRIBE functions in different cells, select all of them, and use the "Run Selection" option in the sidebar to process them in batch.
Processing time depends on the length of the audio file and current API load. Short clips (under a minute) usually process in seconds, while longer recordings may take several minutes.
Yes, if you use the output_doc_name parameter to create a Google Doc, you can edit the document directly. If the transcription is placed in a cell, you can edit it like any other cell content.
The basic transcription will capture all speech but won't distinguish between speakers. For speaker diarization (identifying who said what), you can use an additional TASK_LM step to format the output.
1 week ago