Uploading PDFs and screenshots directly to Claude burns through your usage limit fast. A single PDF page costs 1,500–3,000 tokens, and one screenshot can use up to 4,784 tokens. Converting files to plain text or markdown first can cut your token usage by over 90%.
What PDFs and Screenshots Actually Cost
When you upload a PDF to Claude, it doesn't just read the text. According to Anthropic's documentation, the system converts each page into an image and extracts the text alongside it. That means every single page costs you both image tokens and text tokens — typically 1,500 to 3,000 tokens per page.
A 10-page PDF? That's up to 30,000 tokens gone in one upload.
Screenshots are even worse. According to Anthropic's vision documentation, a single high-resolution image on Opus can use up to 4,784 tokens. If you're screenshotting text and uploading it instead of pasting the text directly, you're wasting tokens on pixels Claude doesn't need.
The Fix: Convert Before You Upload
The solution is simple — don't upload the raw file. Convert your PDFs and screenshots to plain text or markdown before you paste them into Claude. Text uses a fraction of the tokens that images and PDFs consume.
Most people don't realize Claude is processing their PDF as images behind the scenes. You think you're uploading a document, but Claude sees pictures of pages. That's why a 5-page PDF can eat more of your limit than a 2,000-word conversation. The fix takes 10 seconds and saves you hours of waiting for your limit to reset.
How to Do It (Free Tool)
I built a free converter that handles this for you:
- Go to drerinjacques.com/converter
- Upload your PDF or screenshot
- Copy the extracted text or markdown
- Paste it into Claude instead of the original file
That's it. Same content, a fraction of the token cost.
Every PDF page and screenshot you upload to Claude is costing you 5–10x more tokens than necessary. Convert your files to text first using a free tool like drerinjacques.com/converter, and you'll stretch your Claude usage limit dramatically.
Sources
- Anthropic — PDF Support Documentation — confirms each PDF page is converted to an image + extracted text, costing 1,500–3,000 tokens per page
- Anthropic — Vision Documentation — confirms high-resolution images can consume up to ~4,784 tokens on Opus models
