HEIC to DOCX decodes the HEIC through the imaging sidecar, OCRs the PNG pixels, then uses our document engine to build an editable Word file.
From real runs
| Element | Result | What happens |
|---|---|---|
| HEIC preprocessing | Changed | The HEIC never went directly to Azure; it was decoded to PNG first, which resolved the earlier HTTP 400 failure. |
| Editable transcript | Kept | The Word file contained selectable text for the heading, paragraph, bullets, owner, due date, and status. |
| Heading style | Changed | The source screenshot title became a document heading, but the original font, size, and card placement were not reproduced. |
| Bullet list | Changed | The three instructions became separate Word paragraphs in the extracted XML rather than pixel-matched bullet rows. |
| Color and bit depth | Lost | The tested HEIC reported 8-bit sRGB with no alpha, and DOCX output kept none of the HEIC color metadata. |
| Callout labels | Changed | Owner, Due, and Status survived, but the DOCX text extraction showed those labels running together without tab stops. |
| Runtime | Kept | The HEIC-to-DOCX request returned a 10.7 KB Word file in 5.5 seconds and consumed one Deep Read credit. |
The HEIC never went directly to Azure; it was decoded to PNG first, which resolved the earlier HTTP 400 failure.
The Word file contained selectable text for the heading, paragraph, bullets, owner, due date, and status.
The source screenshot title became a document heading, but the original font, size, and card placement were not reproduced.
The three instructions became separate Word paragraphs in the extracted XML rather than pixel-matched bullet rows.
The tested HEIC reported 8-bit sRGB with no alpha, and DOCX output kept none of the HEIC color metadata.
Owner, Due, and Status survived, but the DOCX text extraction showed those labels running together without tab stops.
The HEIC-to-DOCX request returned a 10.7 KB Word file in 5.5 seconds and consumed one Deep Read credit.
We converted our HEIC/HEIF sample (115 KB) with AI Deep Read under AI Deep Read on 2026-09-11. It took 4.9 s and produced a 11 KB Word DOCX file, 90% smaller than the input. The engine raised 1 warning, listed below.
Input · sample.heic · 115 KB
1240 × 1754 px, heic
Output · sample.docx · 11 KB
Word document: 23 paragraphs, 3 headings, 1 table, 0 embedded images; fonts Consolas
HEIC-to-DOCX spends 1 Deep Read credit per image after the imaging sidecar decodes it to PNG. The OCR text is then converted to DOCX, so 10-bit color, ICC data, and photo framing do not survive as Word objects.
sips + Tesseract + Pandoc
sips -s format png photo.heic --out photo.png
tesseract photo.png stdout --psm 6 > photo.txt
pandoc photo.txt -o photo.docxThe verified chain creates a DOCX, but it does not infer the same heading and table structure as the Deep Read transcript.
The HTTP API and the MCP server run the exact engine this page uses (AI Deep Read); the response carries the same warnings the web converter shows. Check support first with GET /api/v1/pairs?from=heic&to=docx.
curl -X POST https://filetypeconverters.com/api/v1/convert \
-H "Authorization: Bearer $FTC_API_KEY" \
-F "file=@input.heic" -F "from=heic" -F "to=docx" \
-F "deepRead=true" \
--output output.docxDeep Read needs a model-readable image payload, and the app now uses the imaging sidecar for HEIC decoding. Our document engine builds the DOCX only after OCR text is available.
No. This route is for text extraction, not image preservation. High-bit-depth HEIC color is reduced during the decoded image stage, and the DOCX carries reconstructed text.
No. The tested DOCX contained editable text and no original HEIC metadata. Store the source HEIC separately if capture time, lens data, or color profile matters.