Images, video & files

Ctrl-V pastes an image straight from the clipboard. Paste a file path and Gray attaches it by type: images are downscaled before sending, PDFs arrive as text (via pdftotext, when installed), and video arrives as a still of its first frame. Audio has no portable wire path on OpenAI-compatible providers, so Gray says so instead of silently dropping it.

gray view

Bash output is text, so the agent cannot look at a PNG by printing it. gray view closes that gap: inside a session the bash tool claims gray view <path> — and cat image.png — and attaches the file as an image the model can see.

$gray view plot.png shot.jpg          # png jpg jpeg gif webp bmp heic heif
$gray view demo.mp4                   # a tiled contact sheet of sampled frames
$gray view demo.mp4 --frames 32       # up to 64 tiles (default 16)
$gray view demo.mp4 --native          # send the video itself (Gemini-class models only)

Images are downscaled to 2000 px. Run gray view yourself and it draws the image inline on terminals that speak the Kitty graphics protocol (Kitty, Ghostty, WezTerm).