Document pack for AI uploads
Split PDF, TXT and Markdown into source-tagged Markdown chunks within an exact selected token budget, then download a ZIP.
Prepare documents
Files are processed on your device. Names and contents are excluded from analytics and cleared on reload.
Result preview
No result yet.
Download Markdown ZIPInstructions and example
Select minutes.txt and cl100k_base with a 2,000-token limit. Each Markdown file includes its source header within the 2,000-token limit. PDF chunks retain page numbers, and manifest.json records each chunk’s token count. The tool does not summarize or rewrite content. The separate public example button demonstrates the workflow without a private file.
Frequently asked questions
Are images and table layout preserved?
No. Only PDF text layers are extracted. Columns and table reading order may change; images, scans and attachments are excluded. Blank text pages are skipped, so compare against the source.
Do all AI services count the same tokens?
No. The full exported Markdown is counted with the selected o200k_base or cl100k_base encoding. Other encodings, system instructions, message wrappers and output reserves are excluded. Leave room within your service’s actual limit.
Are files sent to an AI automatically?
No. Processing and ZIP creation happen in your browser. The ZIP includes input filenames and extracted content: review private information before uploading elsewhere. Limits are 10 files, 25MiB, 200 total sections and 200,000 characters, with 100 pages per PDF.
Calculation criteria and reference materials
- OpenAI — tiktoken
Reference for encoding-specific BPE token counts. This tool uses the selected js-tiktoken encoding.
Verification of calculation criteria: · Calculation·verification principles · Report errors