Budget for images you can publish

“Twenty product images” leaves too much open. Twenty first drafts, twenty approved originals and twenty crops ready for upload are different quantities. State the number of products, colour variants, views and delivery formats. Also state what counts as acceptance: correct garment detail, consistent direction, usable framing and no unresolved product questions.

A hypothetical brief might be five adult apparel products, one colour each, with three approved views per product. That means 15 accepted originals. Two additional crops of each original would create 30 derivative files, but they would not create 30 new views of the garment. Count generation, finishing and delivery separately so the estimate does not inflate either the work or the value.

Set a minimum useful delivery too. If only two views of each product pass review, will you release those ten images or wait for the full set? Decide before the deadline. Otherwise the team may keep generating to satisfy a file count when a smaller, accurate gallery would serve the customer better.

The first is preparation: collecting sources, checking versions, photographing missing details and writing the brief. The second is production: photography, generation, edits or retouching. The third is review: comparing candidates with the garment references and consolidating decisions. The fourth is delivery: crops, filenames, upload, descriptions and checking the live page.

Count your own time even if you do not invoice it. Use an internal hourly value consistently across the alternatives you compare. A founder spending an evening repairing poor inputs has incurred a cost even when the generation tool's bill is small. Conversely, a reusable source set or approved direction can make later batches easier; record that benefit when it actually happens.

For Glammo AI, the current billing configuration uses one credit per image or edit and five credits for a four-second video. Paid plans include downloads and commercial use. Check the current pricing page before estimating a real job, and budget custom models or scenes only if the chosen plan includes them. Plan credits are a capacity allowance, not a promise that every generated draft will be accepted. Check current Glammo plans and credit costs

Suppose your target is 15 accepted images. You allocate $30 of tool cost to the project and spend two hours preparing sources and one hour reviewing and delivering files. At an assumed internal time value of $25 per hour, the total is $30 + (3 × $25) = $105. If all 15 required images are accepted, that is $7 per accepted image. These figures are illustrative inputs, not a Glammo quote or a reported customer result.

Now suppose you finish with only 12 accepted images after the same spend. The cost so far is $105 ÷ 12 = $8.75 per accepted image, and the 15-image brief is still incomplete. If another hour of review and $10 of additional production gets the remaining three accepted, total cost becomes $140, or about $9.33 per accepted image. The extra work matters more than the original per-draft price.

You can also plan capacity using an explicit acceptance assumption. If you assume three out of four generated candidates will be usable, 15 accepted images would suggest 20 candidates. That 75% assumption is not a benchmark or guarantee. A small pilot should replace it with your own evidence, and even an average acceptance rate cannot guarantee that each product receives every required view.

A generation that fails technically and a delivered image you decide not to use are different events. Glammo returns credits for failed generations; a completed image that does not meet your creative or garment-accuracy standard still needs to be accounted for. Do not build a budget around the idea that deleting a disappointing draft makes it free.

Review early enough to stop a bad direction. If the first few outputs repeatedly hide the same detail, change the pose, input or brief instead of completing the batch on autopilot. After delivery, retain the actual preparation time, acceptance count and causes of rework. Those three observations make the next estimate more useful than a larger spreadsheet full of guessed precision.