TrOCR vs dots.ocr
Pick dots.ocr for full pages and PDFs: it returns layout, text, tables as HTML and formulas as LaTeX in one pass, and it handles multiple languages. Pick TrOCR only when you already have cropped single text-line images, mostly English handwriting or receipt lines, or you want a recognizer to fine-tune. The main split is scope: TrOCR reads one line, dots.ocr parses a whole document.
TrOCRmicrosoft/trocr | dots.ocrrednote/dots-ocr | |
|---|---|---|
| Type | Open source | Open source |
| Price | 1 credit / page | 2 credits / page |
| Per 1,000 pages | $1.00 | $2.00 |
| Typical speed | 1.2 s / page | 1.5 s / page |
| Best for | Handwritten text lines | Multilingual document parsing |
| Inputs | images | PDF, images |
| Languages | English | 100+ languages |
| Max pages per request | 100 | 500 |
| Tables | No | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | No | ✓ Yes |
| Bounding boxes | No | ✓ Yes |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | No | ✓ Yes |
| On OCRRouter | Available | Available |
Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.
Which to choose
Choose TrOCR if
- It reports 2.89 character error rate on the IAM handwriting set with the large model, and it has separate checkpoints for handwritten (IAM), printed (SROIE) and scene text.
- It is a good base for fine-tuning a recognizer on your own line images.
- It loads in a few lines with Hugging Face transformers under an MIT license.
Choose dots.ocr if
- It takes a whole page and returns layout boxes, categories and text, with tables as HTML and formulas as LaTeX.
- It is multilingual and outputs JSON, Markdown, HTML, LaTeX and bounding boxes, so PDFs convert to Markdown with layout.
- The v1.5 model (dots.mocr) scores 83.9 on olmOCR-bench, up from 79.1 for the original, and a dots.mocr-svg variant converts charts and diagrams to SVG.
Key differences
| TrOCR | dots.ocr | |
|---|---|---|
| Input it reads | Single text-line images only; you need a separate line detector | Full pages and PDFs, with layout detection built in |
| Output | Plain text | JSON, Markdown, HTML, LaTeX, bounding boxes, SVG |
| Languages | Checkpoints are English-focused; other languages need fine-tuning | Multilingual |
| Model size | Base 334M, large 558M parameters | 3B parameters (1.7B LLM backbone) |
| Hardware and speed | Runs locally through Hugging Face transformers; slow versus CRNN recognizers on CPU | Single CUDA GPU with vLLM is the documented path; no VRAM figure stated |
| Benchmark | 2.89 CER on IAM handwriting (TrOCR-large, per the paper) | 79.1 on olmOCR-bench (original), 83.9 (dots.mocr) |
| Known failure | No layout, table or page-level output | Runs of ellipses or underscores can make output repeat endlessly; a table above 1000 grid cells drops the whole page in Docling |
Sources
- TrOCR paper (arXiv 2109.10282)
- trocr-large-handwritten model card
- Transformers TrOCR docs
- dots.ocr GitHub README
- Docling issue #5146
What each one is
TrOCR
Transformer text recognition model. Reads single text lines, including handwriting.
TrOCR detailsdots.ocr
Single vision model from rednote that does layout detection and content recognition together.
dots.ocr detailsTry both on your own file
Run the file once with each model and compare the output.
Use one, fall back to the other
{"model": "microsoft/trocr","fallbacks": ["rednote/dots-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
TrOCR alternativesdots.ocr alternativesBest OCR models in 2026