TrOCR vs Chunkr
Pick TrOCR if you need to read single lines of English handwriting or receipt text, or want a recognizer you can fine-tune on your own line images. Pick Chunkr if you need to parse whole PDFs, slides, Word files or images into HTML, Markdown or JSON with bounding boxes, for example to chunk documents for RAG. The main reason: TrOCR only reads one text line at a time and has no layout output, while Chunkr handles full documents.
TrOCRmicrosoft/trocr | Chunkrchunkr/chunkr | |
|---|---|---|
| Type | Open source | OCR API |
| Price | 1 credit / page | 14 credits / page |
| Per 1,000 pages | $1.00 | $14.00 |
| Typical speed | 1.2 s / page | 3.0 s / page |
| Best for | Handwritten text lines | Layout-aware chunks for RAG |
| Inputs | images | PDF, images, DOCX |
| Languages | English | Multilingual |
| Max pages per request | 100 | 1,000 |
| 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, per the paper.
- It ships separate checkpoints fine-tuned for handwritten (IAM), printed (SROIE) and scene text, and loads in a few lines with Hugging Face transformers.
- It is MIT licensed and free to self-host, so you pay only for compute and have no copyleft obligations.
Choose Chunkr if
- It takes PDF, PPT, Word and images and returns HTML, Markdown, JSON and OCR with bounding boxes, so you do not need a separate line detector.
- It offers both a cloud API and a self-host version, with Docker options for GPU, CPU-only and Mac ARM.
- It works with any OpenAI-compatible LLM endpoint and is suited to RAG chunking of PDFs and slides.
Key differences
| TrOCR | Chunkr | |
|---|---|---|
| Input scope | Single text-line images only; you need a separate line detector | Whole documents: PDF, PPT, Word and images |
| Output | Plain text | HTML, Markdown, JSON and bounding boxes |
| Layout and tables | No layout, table or page-level output | Layout analysis with semantic chunking |
| Licence | MIT | AGPL-3.0 or commercial license; closed-source use needs the commercial license |
| How you run it | Library, run locally through Hugging Face transformers | Cloud API or self-host with Docker; self-host needs an external LLM endpoint |
| Price | Free, self-hosted (you pay for compute) | Hosted credit-based plans, prices not published; open-source version is free, self-hosted |
| Benchmark | 2.89 CER on IAM handwriting (TrOCR-large) | Not published |
Sources
- TrOCR paper (arXiv 2109.10282)
- trocr-large-handwritten model card
- Transformers TrOCR docs
- GitHub README
- Chunkr site
What each one is
TrOCR
Transformer text recognition model. Reads single text lines, including handwriting.
TrOCR detailsChunkr
Document parsing API that segments and chunks pages for RAG. Self-hostable version available.
Chunkr 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": ["chunkr/chunkr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
TrOCR alternativesChunkr alternativesBest OCR models in 2026