CartVisualCreate account

How does Claid’s default four-image batch affect credits in Image Generation and AI Background?

Both endpoints document four candidates by default, but the credit rate differs. Image Generation charges one credit per generated image, while AI Background documents two credits per generated image. Set options.number_of_images for Image Generation or output.number_of_images for AI Background to fit your review. Tune Image Generation’s guidance_scale separately when prompt adherence needs adjustment.

Published · By CartVisual editorial team

Separate the two controls

number_of_images answers a production question: how many candidates should one request return? guidance_scale answers a creative-control question: how closely should the outputs follow the prompt? They should not be increased together automatically. More candidates expand the search space; stronger guidance narrows the prompt interpretation, with a possible effect on quality and diversity.

The practical decision is therefore a pair, not a single “quality” setting. Decide the candidate budget first. Only then tune prompt adherence for that budget. This prevents a team from silently multiplying spend while trying to solve a prompt problem.

  • Use number_of_images when the main uncertainty is which composition or interpretation will work.
  • Use guidance_scale when the wording is already settled but the output needs to follow it more closely.
  • Treat the source documentation’s suggested values as starting points, not a guarantee for a particular SKU or prompt.

Calculate exposure before sending

Image Generation documents a range of one to four candidates and charges one credit per generated image. Four candidates cost four credits at that documented rate. Set options.number_of_images in the Image Generation request.

AI Background accepts one product image per request and documents a separate output.number_of_images control. Its candidates are 1024 × 1024, and the documented charge is two credits per generated image. The four-candidate default therefore represents eight credits per request. Output encoding and temporary storage are separate delivery settings.

EndpointCount fieldDefault planned costOne candidate
Image Generationoptions.number_of_images4 × 1 = 4 credits1 credit
AI Backgroundoutput.number_of_images4 × 2 = 8 credits2 credits

Use a controlled comparison

The documented options.guidance_scale range is 1.0–49.99, with a default of 5.0. Claid notes that values between 5 and 7 usually provide good results, while 12 may offer greater precision. It also warns that changing the scale may affect quality and diversity. Do not interpret 12 as a guaranteed improvement; it is a different control setting that requires review.

A useful planning method is to hold one variable still while examining the other. If the prompt is stable, compare candidate counts at the same guidance value. If the candidate count is fixed, compare two guidance values and inspect whether the result better follows the brief without losing useful variation. Record the actual setting beside each output, rather than relying on a remembered default.

For ecommerce use, approval should concern the returned asset and the authoritative product source. A setting that follows a scene description closely does not establish that product details, dimensions, packaging, or text are correct. Generated imagery remains a creative candidate until reviewed.

Illustrative credit plan for three products

A shop needs one prepared background candidate for each of three product photos. With three AI Background calls, the default generates 12 candidates for 24 credits at the documented rate. Setting output.number_of_images to 1 gives three planned candidates for six credits.

Separately, the shop plans one Image Generation request to explore a new scene. Two candidates represent two credits. If its prompt is settled, start with the documented guidance default and review adherence before changing the value. Record the endpoint, exact field, candidate count, and intended credits before submitting.

Related questions

Does choosing four images mean the request is better than choosing one?

No. It increases the number of candidates and, according to Claid’s documentation, the corresponding per-image credit exposure. Choose four only when the review benefits from additional alternatives.

Where CartVisual fits

CartVisual can help create product-image candidates using its own supported workflow, but this route is not a Claid API integration. Review any generated candidate against the authoritative product source before publishing.

See the current CartVisual workflow and its limitations

Source and evidence scope

Prepared by the CartVisual editorial team with AI assistance. Competitor behavior is documented rather than independently executed for this guide.

Claid: Image Generation Options

Official source snapshot checked .

Claid: AI Background I/O

Official source snapshot checked .

The two Claid endpoint references establish count, default and per-image credits. AI Background documents one-product input and square output. Only the Image Generation reference here establishes guidance_scale. The examples calculate planned usage from the published rates.

Related decisions

Browse ecommerce creative workflows