Understand Images, Video, and Audio

Understanding actions select a strong default model automatically. Specify --model only when you need a particular model.

Read images

Analyze up to 10 local files and URLs in one request:

anycap actions image-read \
  --file ./before.png \
  --file ./after.png \
  --instruction "List every visible change" \
  | jq -r '.content'

Read video

anycap actions video-read \
  --file ./demo.mp4 \
  --instruction "Summarize the flow and list all on-screen text" \
  | jq -r '.content'

Direct video URLs and public YouTube URLs are also supported. Exactly one --url or --file is required. Local video files must be no larger than 100 MB.

Read audio

anycap actions audio-read \
  --file ./meeting.mp3 \
  --instruction "Transcribe the decisions and action items" \
  | jq -r '.content'

Direct audio URLs and YouTube URLs are supported.

Response shape

{"status":"success","content":"...model response...","request_id":"req_abc123"}

Local-file rule

Pass local files directly with --file. Do not upload them to Drive first. The action handles its own upload and returns only the analysis text and diagnostics.