TrOCR Base
Free.ai (self-hosted)
·
ocr
·
~300 tokens per page
TrOCR Base is an OCR model built by Microsoft. Strongest at Handwritten documents, low-resource scripts.. Self-inotarisirwa pa Free.ai GPUs - inofamba zvakasununguka pazuva rako re token pool (300 tokens papeji). Yakaburitswa pasi peMIT - kushandiswa kwekutengesa kwakabvumirwa pasi peFree.ai.
Usati washandisa
OpenAI-inowirirana REST API. Kugadzira chiratidzo uye kukumbira iyi model mumasekondi.
curl -X POST https://api.free.ai/v1/ocr/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"trocr-base","prompt":"your prompt here"}'Zvimwe zvinobvunzwa kakawanda
TrOCR Base anotora mazita kubva kune mapikicha uye PDFs. Upload a scanned page, a screenshot, kana a photo of a document and TrOCR Base returns the text - preserving line breaks and (on capable engines) tables, formulas, and structure.
TrOCR Base inobata mazhinji emazita ezvinyorwa - Latin, Cyrillic, CJK, Arabic, Devanagari, uye zvimwe. / ocr / landing ine yakazara runyorwa rwemazita ezvinyorwa.
Yes for clean printed-style handwriting; cursive and shorthand are harder. TrOCR and modern transformer-based engines are noticeably better than legacy Tesseract on handwriting.
Layout-aware engines (GOT-OCR, Marker, Florence-2) idzoreredza Markdown/HTML mapepanhau uye LaTeX for formulas. Plain-text engines idzoreredza mapepanhau pasina hunhu — tarisa izvi model's output mode.
TrOCR Base inoita paGPU yedu kubva pazuva nezuva pasina pool; $5 → 200,000 paid tokens after. ~300 tokens per page.
PNG, JPG, WebP, HEIC, BMP, pamwe nePDFs ine imwe uye yakawanda peji kusvika ku50 MB. Kukanganisa kana kutenderera peji kunogadziriswa otomatiki mushure mekubvisa.
Character accuracy inowanzoita>98% pazvinyorwa zvakadhindwa zvakachena, kudonha pazvinyorwa zve low-res, zvakaoma kusvibiswa, kana zvisina kujairika fonts. Compare engines on /ocr/compare/ when accuracy matters.
Yes - /batch/ inogamuchira folder yemafaira emafaira/PDFs. Each extraction lands in /account/?tab=history. For folder-tree preservation use the API.
Yeah — POST yako faira kuti /v1/ocr/ with model="TrOCR Base". Returns JSON with text + (kana yakatsigirwa) layout + word-level bounding boxes. /api/ ine yakazara reference.
Self-inochengetwa mamodheru kuchengetedza faira rako paGPUs yedu; premium kutenderera kuburikidza neDPA. Mapepa anoparadzwa mushure mekugovera-window. Isu hatina kudzidzisa pane yako inputs.
Yeah — Free.ai inobvumidza kushandiswa kwezvigadzirwa zvemapepa. Unofanira kunge uine kodzero dzekushandisa chinyorwa chemutauro wechigadzirwa.
1-5 masekondi papeji zvakajairika. Layout-aware injini (tabs, math) kutora nguva yakareba. Multi-page PDFs kuchinja zvakaenzana linearly.