Skip to content
AnglinAI Docs

ERR_CONVERSION_EMPTY_OUTPUT — Conversion Empty Output

HTTP Status: 500 Retryable: Yes Automatic retry: None. However, retrying with a different quality tier may produce different results.

What the User Sees

The conversion produced no output. This may happen with scanned or image-only PDFs. (ERR_CONVERSION_EMPTY_OUTPUT)

What Causes This Error

The AI model returned valid responses but the extracted HTML content is empty or trivially small (below a minimum character threshold). The system detects this condition in routes/convert.ts after all conversion steps complete.

Common causes include: scanned documents where OCR fails to extract text, image-only PDFs with no text layer (photographs, artwork, hand-drawn diagrams), heavily stylized documents with unusual fonts or text rendering that confuses the vision model, and extremely low-resolution scans where text is not legible even to AI vision models.

When this error occurs, the system sends both an admin notification email and a user notification email with details. Credits are NOT deducted for empty conversions — the file is marked as failed with autoRefunded: true.

This is one of the few errors where retrying with a different quality tier can make a meaningful difference, as different models have different strengths with OCR and image recognition.

Resolution Steps

For Users

  1. Try the “premium” quality tier, which uses more sophisticated vision models that may handle difficult documents better.
  2. Ensure the PDF has a text layer. Try selecting text in a PDF viewer (Adobe Acrobat, Chrome) — if you cannot select text, the PDF is image-only.
  3. If the PDF is a scan, use dedicated OCR software (Adobe Acrobat Pro, ABBYY FineReader, or online OCR tools) to add a text layer before converting.
  4. Verify the PDF content is legible — very low-resolution scans or heavily degraded copies may not produce usable output.
  5. Your credits were not deducted for this failed conversion.

For Administrators

  1. Check the conversion failure alert email for details including token usage, cost, and file ID.
  2. The file is marked as failed with autoRefunded: true in the metadata, confirming credits were not charged.
  3. Review the pipeline log for the failed conversion via Admin > Files. Check which chunks (if any) produced output and which were empty.
  4. If this error is frequent for a specific document type, consider adding pre-conversion checks to detect image-only PDFs and warn users before they attempt conversion.
  5. Monitor the error rate in Grafana. A sudden increase may indicate a regression in the AI model’s OCR capabilities.
  6. The minimum output threshold is defined in routes/convert.ts — adjust if the threshold is too aggressive.