Przejdź do treści

Examples

Three worked sessions with an AI assistant connected to Qamera AI — from an empty workspace to a packshot, a reviewed set, and a short video.

These examples show the shape of a working session: what you ask, which tool the assistant reaches for, and what comes back. Your assistant will phrase things differently, and that is fine — what matters is the order of the steps.

Identifiers are shown as they arrive. Copy them unchanged; they are not meant to be readable or reformatted.

Example 1: from an empty workspace to a packshot

The first thing to do in a new workspace is put a product in it. Nothing can be generated before that.

You: "Here are three photos of a linen shirt. Add them to Qamera as one product called 'Linen shirt, ecru' and then make me packshots."

Assistant: calls register_product with the three URLs, name: "Linen shirt, ecru", productType: "shirts"

Registered as one product, id 9f1c2b64-…. The platform is analysing the photos now — I can't order generation until that finishes.

Assistant: calls get_products and checks the status

Status is DESCRIBED, so it's ready.

You: "Ten packshots, on a plain background."

Assistant: calls get_presets to see what styles are available, then register_packshots with productId and count: 10

Ordered. Order id 4d8e1a07-…. This spends credits and runs in the background.

You: "Are they done?"

Assistant: calls check_generation_status with the order id

completed — ten of ten jobs finished, no errors.

Assistant: calls get_packshots with orderId so only this order's results come back

Ten packshots, ids listed. Want me to look through them?

What to notice. The assistant filtered get_packshots by orderId. Without it the list returns everything in the workspace, newest first, and you would be reading results from older sessions too.

Example 2: reviewing before you approve

Lists give the assistant identifiers, not pictures. To form an opinion it has to open each one.

You: "Go through the packshots from that order and tell me which are worth keeping."

Assistant: calls get_packshots with the orderId, then get_packshot_preview once per packshot

I've looked at all ten. Seven are clean — product centred, even background, no clipping at the edges. Two have a visible shadow along the bottom edge. One is cropped tighter than the rest, so it will not sit well next to the others on a listing.

You: "Approve the seven, reject the rest."

Assistant: calls update_packshot for each, with voting: "APPROVED" or voting: "REJECTED"

Done — seven approved, three rejected. The rejected ones stay in the workspace; say the word and I'll archive them with status: "ARCHIVE".

What to notice. The assistant can only comment on packshots it actually opened with get_packshot_preview. If it offers an opinion on a list it never previewed, ask it to preview them first.

A word about wording. Packshots are voted APPROVED / REJECTED. Photo-shoot images use accepted / rejected instead, through update_image. The two are separate and your assistant has to use the right one for the right kind of image.

Example 3: a short video from an image you like

Video is generated from one image you already have, not from a product.

You: "Run a photo shoot for the linen shirt — outdoor, daylight, four images, portrait."

Assistant: calls get_sceneries, then register_photo_shoot with productId, sceneryId, aspectRatio: "4:5", count: 4, and suggestions: "outdoor, natural daylight"

Ordered, order id b70f5c93-…. aspectRatio is required for a photo shoot, so I chose 4:5 for portrait — tell me if you'd rather have something else.

You: "Show me the results."

Assistant: calls get_images with the orderId, then get_image_preview for each

Four images. The second one has the best light on the fabric.

You: "Make a short video from that one, slow zoom."

Assistant: calls register_video with generatedImageId and `suggestion: "slow zoom in, product stays centred"

Ordered. I'll check get_videos in a few minutes — video takes longer than stills.

You: "Ready?"

Assistant: calls get_videos

Ready, and the file address is in the answer.

Getting more out of it

  • Be specific about numbers and format. "Ten packshots" and "portrait, 4:5" reach the tools as parameters. "Some images" does not.
  • Ask for a preview before a verdict. An assistant that has not opened an image cannot judge it.
  • Name a preset when consistency matters. Presets are what keep a second batch looking like the first.
  • Watch your credits. Every register_… step spends them; reading, previewing and voting do not.
  • Say which order you mean. Filtering by orderId is the difference between this session's results and everything you have ever generated.