AI Zoom virtual background image: 16:9, 1280x720, PNG or JPG

Zoom asks for a 24-bit PNG or JPG under 15 MB, at least 1280x720 pixels, cropped to your camera's aspect ratio; a 16:9 camera fits 1280x720 or 1920x1080. On Sume, request aspect_ratio: "16:9" from a model that lists it, then check the returned pixel size before you upload. A 16:9 render costs between $0.025 and $0.10 per image on the catalog rows below.

What Zoom's page says

Figures from Zoom's virtual background support page (read 2026-10-03).

Pick a Sume model

Models that list 16:9 and take no reference are enough for a room scene. The Image API docs say a model only accepts the values its catalog descriptors list, so read supported_parameters first. These rows list 16:9 and the price shown is the endpoint line you are charged (read 2026-10-03):

• Zoom's 1280x720 floor is a pixel count, and the docs do not promise a fixed output size for every row. Open the file, read its width, and regenerate on a row with a higher tier if it falls short.
• Nano Banana 2 is the one row here whose descriptor lists resolution; send resolution: "2K" when you need headroom.

Compose for the camera, not the wall

A virtual background sits behind a person who fills the middle third of the frame. Ask for the interesting detail on the left and right and leave the center calm. Avoid readable text in the picture: image models can still struggle with precise text, per OpenAI's image guide, and a mirrored or garbled sign behind you looks worse than none.

Make a small set rather than one: a home office, a plain studio wall and a bookshelf cost about $0.11 together on Flux 2 Pro, and you can pick the one that sits best behind you. Prompt example: 'a quiet home office, wide 16:9, shelf with plants on the left, window on the right, empty soft wall in the center, natural light, photo'.

A request you can run

Replace the model with any row from the table. This writes a JPEG; check the size before you upload.

curl -s https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "black-forest-labs/flux.2-pro",
    "prompt": "a quiet home office, shelf with plants on the left, window on the right, empty soft wall in the center, natural light, photo",
    "aspect_ratio": "16:9",
    "output_format": "jpeg"
  }' | python3 -c "import sys,json; print(json.load(sys.stdin)['data'][0]['url'])"

What Sume does not do

Sume does not check your camera's ratio or cut a person out. If your camera is 4:3, ask for 4:3 instead, which every row in the table also lists. A 200 response carries the image; a 202 means the job is still running and you read it from the job result endpoint.