Surya
App in the BluixApps catalog
What it is
Surya OCR is Datalab's modern document AI toolkit — multilingual OCR (90+ languages), layout analysis, reading order detection, and table recognition in one package. Significantly higher accuracy than Tesseract on real-world documents (magazines, forms, scanned photos).
The 2024 generation of document AI, the canonical alternative to Tesseract for modern OCR workflows.
What it's for
- Multi-language OCR — 90+ languages
- Layout analysis — section blocks (title, paragraph, table, figure)
- Reading order detection — correct text flow on complex pages
- Table recognition — extract structured tables
- Form processing — extract key-value pairs
- Document classification — by content type
Who it's for
- Document AI teams processing real-world inputs
- Legal / contract platforms OCRing scanned documents
- Operula digitizing artisan documentation, certificates
- Invoice / receipt processing workflows
- Academic researchers processing historical documents
- Hosting providers offering document AI tier
Why teams pick Surya OCR over alternatives
- GPL-3.0 — fully open
- Better than Tesseract on modern documents (forms, magazines, screenshots)
- Built-in layout + table — Tesseract requires plugins
- 90+ languages — broad coverage
- Active maintenance — Datalab continuous improvements
- Streamlit UI included for non-technical users
- API-friendly for batch processing
Integrations
- Streamlit web UI (BluixApps default launcher)
- Python API for batch processing
- CLI mode for command-line workflows
- Pair with: NLLB-200 (OCR → translate)
- Pair with: LLM (OCR → entity extraction → structured data)
- PDF + image input formats
- Outputs: JSON, Markdown, CSV (for tables)
Notable users & community
- 10k+ GitHub stars
- Datalab + extensive contributor base
- Featured in document AI roundups as Tesseract successor
- Active research integration with modern LLM workflows
- Multiple commercial integrations
Tips & operations
- Languages:
- All EU languages
- Chinese, Japanese, Korean, Vietnamese
- Arabic, Hebrew, Persian, Urdu
- Indian languages (Hindi, Bengali, Tamil, etc.)
- Many indigenous + research languages
- Speed:
- GPU (RTX 4090): 1-3 sec per page
- CPU: 10-30 sec per page
- VRAM: 4 GB minimum
- Pipeline stages:
- OCR: text extraction with bounding boxes
- Layout: classifying regions
- Reading order: correct flow
- Tables: structured extraction
- CLI batch: process entire folders
- Best inputs: scanned PDFs, photos of documents, screenshots
- Surya vs Tesseract:
- Surya: better real-world accuracy, layout-aware
- Tesseract: faster on simple printed text, lower memory
What we ship in BluixApps
- Docker (pytorch CUDA 12.4 + surya-ocr + streamlit + poppler-utils)
- Streamlit GUI launcher (
surya_gui) - Persistent volumes: cache (models, ~2 GB), input, output (JSON/MD/CSV)
- Port 7883 mapped
- Install report at
/root/bluixapps/surya.txt - Language guidance
- Pipeline stage documentation
- Surya vs Tesseract comparison
- Use case examples (legal, archives, invoices)
- Pairing suggestions (NLLB, LLM for entity extraction)
- GPU pre-flight check via
bluixapps_ensure_nvidia_runtime - Backup hook covers cache + output
Get this app — pick a BluixApps plan
Same catalog. Scaling tenant isolation, white-label and support tier.
| Tier | Tenants | Catalog | Support | White-label | Monthly | |
|---|---|---|---|---|---|---|
| Stacks | 1 | 19 curated stacks | Standard | — | $19/mo | DetailDeploy |
| Starter | 10 | Full catalog | Standard | +$15–25/mo | $49/mo | DetailDeploy |
| Pro | 25 | Full catalog | Priority bugfix | +$15–25/mo | $149/mo | DetailDeploy |
| Growth | 100 | Full catalog | Priority bugfix | +$15–25/mo | $349/mo | DetailDeploy |
| Scale | 500 | Full catalog | 7-day window | +$15–25/mo | $799/mo | DetailDeploy |
| Enterprise | Unlimited | Full catalog | Priority 7-day | Bundled | $1,499/mo | DetailDeploy |
