PDF to Markdown
상업적 사용 OK
380+ 모델
워터마크 없음
가입이 필요하지 않습니다
모델:
+ GPT-5, Claude, Gemini
Drop a PDF — AI converts it into clean GitHub-flavored Markdown with headings, paragraphs, lists, tables, and code blocks all preserved. Powered by IBM Granite-Docling-258M (Apache 2.0). Faster + smarter than plain text extraction.
Extracting layout-aware Markdown… ~5-10 sec/page
고급 옵션
결과
토큰이 부족해요
Get More Tokens
더 먹고 싶어?
하루 5K 토큰 + 10K 보너스 무료 가입
무료로 가입하세요
귀하의 요청을 처리 중...
Convert any PDF into clean GitHub-flavored Markdown with headings, tables, lists, and code blocks preserved. Powered by IBM Granite-Docling. Free, unlimited, no signup.
사용 방법 PDF to Markdown
1
입력을 입력하십시오
텍스트를 입력하거나 파일을 업로드하거나 원하는 내용을 설명하세요. 계정이 필요하지 않습니다.
2
생성하기를 클릭하십시오
당사의 AI는 최고의 오픈 소스 모델을 사용하여 몇 초 만에 요청을 처리합니다.
3
다운로드 및 공유
다운로드, 복사 또는 결과를 공유. 개인 및 상업용 무료.
API를 통해 이 도구를 사용
이 도구를 자신의 코드로 자동화하세요. OpenAI 호환 REST 엔드포인트, 베어러 토큰 인증, 추가 SDK 필요 없음. 토큰 비용은 웹 인터페이스와 일치합니다.
curl -X POST https://api.free.ai/v1/chat/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model": "qwen7b", "messages": [{"role": "user", "content": "Use the PDF to Markdown tool on: ..."}]}'PDF to Markdown — FAQ
Drop in any PDF and the AI converts it into clean GitHub-flavored Markdown — headings stay headings, tables stay tables, lists stay lists, code blocks stay code blocks. Goes way beyond plain text extraction; the document's structural hierarchy is preserved so you can drop the output straight into a docs site, an LLM RAG pipeline, or a search index.
IBM Granite-Docling-258M (Apache 2.0). Tiny vision-to-sequence model fine-tuned for layout-aware document conversion — beats pdftotext + much faster + smarter than running a generic vision-language model on each page.
pdftotext is a flat dump — paragraphs and tables collapse into a wall of words. Adobe Export to Word preserves layout but produces .docx + costs ~$15/mo. Docling preserves the SEMANTIC structure (heading levels, lists as lists, tables as Markdown tables) and outputs a format LLMs and dev tools can both consume natively.
LlamaParse and unstructured both have free tiers but cap pages/month and require an API key. Docling-258M runs locally on our GPU + is fully self-hosted Apache 2.0, no per-page metering, no key signup. Quality is competitive with LlamaParse on standard documents.
Yes — tables come back as proper Markdown pipe-tables. Complex multi-column / nested tables are flattened more aggressively (a fundamental Markdown limitation, not the model's fault). For perfect table fidelity, we also support `format=html` via the API which preserves rowspan/colspan.
Granite-Docling does the OCR step itself — works on born-digital AND scanned PDFs alike. Scanned at lower DPI (<150) loses some text accuracy; rescan at 200+ DPI for best results.
Most LaTeX-rendered equations come through as inline `$...$` Markdown math. For research papers with heavy math, we also offer the academic-paper-extract tool (Nougat) which is specifically tuned for equations and citations.
About 5-10 seconds per page on our H200. A 30-page report is ~3-5 minutes. Tiny model means batches of small PDFs are essentially free in the daily pool.
200 tokens per page, with a 500-token floor. A 5-page contract = 1,000 tokens. A 30-page report = 6,000 tokens. The 5K daily free pool covers most typical use.
PDF — born-digital + scanned both supported. Max 50 MB upload. Other document formats (DOCX, EPUB, HTML, etc.) are on the roadmap; for now upload-and-convert with the pdf-conversion tool first.
Processed immediately, the Markdown output is kept (24h anonymous / 7d paid share-link expiry), the source PDF is deleted right after extraction. Never used for training. /privacy/ for the full policy.
Yes — POST a multipart `file` to /v1/document/pdf-to-markdown/. Returns {markdown_url, pages, preview, tokens, share_url}. Bearer auth (sk-free-…) gives 10K free tokens/month. /api/ has the curl example.
이 도구를 어떻게 평가하시겠습니까?