PDF2Dataset

Convert PDFs and images into clean Hugging Face datasets with Mistral OCR, structure-aware chunking, validation, deduplication, and quality filtering.

Source documents

Chunking

Max chunk size

Character budget per chunk. Use 0 to keep each document as one chunk.

0 4096

Quality filters

Minimum characters

0 500

Minimum words

0 50

Hugging Face output

Dataset repository

Hugging Face token

Append to existing dataset

Pipeline report

Document preview

Uploaded documents

Requires MISTRAL_API_KEY, a Hugging Face token, or an active Hugging Face CLI login.

Examples

Max chunk size	Hugging Face token	Dataset repository	Append to existing dataset	Minimum characters	Minimum words