How do I resolve the OneDrive download node not producing output and split large audio files for OpenAI transcription in n8n?

Below is some guidance based on the documentation and common approaches when working with large audio files and download nodes in n8n.

---

OneDrive Download Node Producing No Output

Because the OneDrive download node ran “successfully” (i.e. it showed the node completed) but did not produce any output, consider the following:

1. File Filtering and Selection
   • Verify that the OneDrive trigger is sending only the file(s) you expect. For example, if multiple files (MP3 and MP4) are arriving, use additional nodes (such as a Sort node or a Limit node) to select only the most recent file.
   • Check if the file metadata (like file size and mime type) can be used to filter out unexpected files.
   If the node was previously used with Google Drive and a smaller audio file worked, it may be that the OneDrive trigger is bringing in additional files so that the download node isn’t “picking up” the intended file.

2. Large File Size Issues
   • The downloaded file (especially an MP4 around 450,000 KB) is much larger than the typical file sizes handled by subsequent operations (for example, by OpenAI’s transcription API, which limits files to 25 MB).
   • Even if OneDrive could download the file data, n8n (especially in the cloud service) might hit processing or memory limitations when trying to pass such a huge binary along the workflow.
   • As a troubleshooting step, try manually limiting the files sent from your trigger (for example, using a Filter or even switching temporarily to a known smaller file) to confirm that the issue is related to file size.

---

Breaking Up a Large Audio File for Transcription

Since the OpenAI transcribe operation (using the Whisper model) accepts audio files up to 25 MB, you must split the large file into smaller parts before transcribing. Here are a few strategies:

1. Split the Audio File Using Workflow Nodes
   • If you have a self-hosted setup, many communities use FFmpeg to cut the audio into smaller segments. Unfortunately, while that’s a common recommendation for self-hosted n8n, on n8n cloud you don’t have direct access to custom FFmpeg nodes.
   • Instead, you could try:

   • Using a service (or HTTP Request node) that offers audio processing to segment your file.
   • Designing a workflow that first offloads the large file to a service that can split it (or use an external API), then retrieves and passes the smaller files to the OpenAI node.

2. Using the Split In Batches Node
   • While the Split In Batches node is typically used for processing many items by breaking them into smaller groups, note that it does not split a single binary file.
   • Its use case is better suited for cases where you have an array of small audio segments. If you can pre-split your audio (for example, by saving the file in segments), you can then loop over each segment to run transcription on each, and finally merge the results together.

3. Merging Results After Transcription
   • Once each segment is transcribed, you can use nodes like the Merge or Function node to combine the text outputs back into one coherent transcription.
   • For example, accumulate your results in an array and then join them with a newline or appropriate separator.

---

Summary of Actions

• For the OneDrive Issue:
  1. Check that your trigger node and download node are correctly filtering for the intended (most recent) file.
  2. Confirm that the OneDrive download node isn’t silently dropping large files due to memory or file size limits inherent in n8n cloud.
• For Splitting and Transcribing a Large Audio File:
  1. Since OpenAI only accepts files under 25 MB, you’ll need to break up the large file.
  2. If you cannot use FFmpeg on n8n cloud, consider using an external service via an HTTP Request node to segment the audio.
  3. Process each segment individually using the OpenAI transcribe node, and then combine the outputs into the final text.

---

Keep in mind that these approaches are workarounds based on our current documentation guidelines. If you continue to run into issues with large files in n8n cloud, you might need to consider additional external processing services or evaluate moving to a self-hosted option where you have full control over resources and can leverage FFmpeg directly.