PNG, JPG or WebP? The Best Image Format for AI Vision Models
Which image format and resolution give the best results with GPT, Claude and Gemini vision, and how to avoid wasting tokens on oversized images.
Format by content
Screenshots, UI, code, and diagrams: PNG, because lossless edges keep small text readable. Photos: JPG at high quality, because PNG photos are large with no accuracy gain. WebP is fine where supported, but PNG and JPG are universally accepted.
Resolution sweet spot
Vision models downscale large images internally. A 4000-pixel photo is shrunk before the model sees it, and you may pay for tiles it never needed. Crop to the region you care about instead of sending the whole screen.
Text and motion
If you only need the words, extract them with Screenshot to Text and send text, which is cheaper and more precise. For UI interactions or animations, a short GIF made with Video to GIF shows the sequence in one compact file.
Try it free in your browser
- Screenshot to Text — Extract readable text from any image with browser-side OCR.
- Video to GIF for AI Vision — Convert MP4, MOV, or WebM clips to compact GIFs that fit Claude, Gemini, and GPT vision uploads. Runs entirely in your browser.
Tools mentioned
Frequently asked
Does image resolution change cost?
Yes. Most vision models charge by image tiles or pixels, so oversized images cost more without adding detail.
Why does text in my screenshot get misread?
Compression artifacts in JPG blur small text. PNG keeps edges crisp.
Keep reading
How to Prepare PDFs for ChatGPT, Claude, and Gemini
A practical guide to extracting clean, AI-ready text from PDFs — born-digital and scanned — so ChatGPT, Claude, and Gemini answer accurately and don't waste tokens on headers, footers, and page numbers.
The Best Way to Feed Long Documents to Claude (and Other Long-Context Models)
Claude's 200K-token context is generous, but you'll still want to clean, compress, and structure long documents before sending them. Here's a step-by-step playbook.
Cleaning Whisper Transcripts for AI Summaries
OpenAI Whisper, Otter, and YouTube transcripts are full of timestamps, filler words, and speaker noise. Here's how to strip them before sending to ChatGPT or Claude — and why it matters.