Image Alt Text Template Generator

Alt text has two audiences that want the same thing. Someone using a screen reader needs to know what the image conveys; a crawler needs to know what the image is about. Both are served by a short, specific, honest description - and neither is served by a keyword list or by the filename the camera assigned.

This tool builds alt text from templates. Enter a product or page name and optional keywords, and it assembles a set of candidate phrasings by combining them in various orders. A bulk field takes one name per line and produces a first-pass description for each.

The word to hold onto is candidate. These are string combinations, not descriptions - nothing here has looked at your image - and the honest way to use them is as a starting structure that you then correct against what is actually in the picture.

A templated description gets you above the worst outcome, which is no alt attribute at all or an attribute containing "IMG_4471.jpg". For a catalogue of two hundred product shots that all look broadly alike - object, white background, front three-quarter view - a template is very close to what a human would have written, and generating it in one pass is a legitimate use of an afternoon.

It stops being adequate the moment the images differ from each other in ways that matter. A lifestyle photograph of the same product in use, a size-comparison shot, a detail crop of the stitching, and a packaging shot all deserve different descriptions, and giving them the same templated string tells a screen-reader user that four identical images are present when in fact each carries different information.

So the workflow that respects both audiences is: template the bulk, then rewrite the ones that carry unique information. In practice that is usually the hero image, anything containing text, anything where the image is making an argument the surrounding copy does not make, and any diagram or chart.

The bulk field takes filenames or names one per line and returns one description each, using the same first template for every row. That uniformity is the point - it produces a consistent, predictable string you can paste into a spreadsheet column and import, which is how alt text actually gets applied at scale in a product catalogue or a content management system.

What it means in practice is that the keywords you enter apply to the whole batch. That works when the batch genuinely shares attributes: forty photographs of the same product in different colours, or a set of images all belonging to one category. It fails when you paste a mixed list, because "noise cancelling, over-ear, black" will be appended to a description of your office building.

Split the work by batch rather than by tool, then. Group images that share attributes, run each group with its own keywords, and handle the one-off images individually. That is more passes through the tool and considerably less rewriting afterwards.

Five base phrasings are built from the name alone - the name followed by "product photo", the name followed by "image", "Photo of" plus the name, the name with a high-quality-product-image suffix, and the name on a white background. These exist so that a name with no keywords still yields a usable set.

Your keywords are then split on commas, semicolons, and newlines, trimmed, and emptied of blanks. For each keyword, three further phrasings are added: the name joined to the keyword with a dash, the keyword joined to the name with a dash reversed, and the name and keyword followed by "product photo". If there are two or more keywords, one additional phrasing joins the first two with "and". The whole list is then deduplicated, so a keyword that happens to repeat a word in the name does not produce two identical entries.

The bulk column runs the same builder for each line and takes the first result - the "product photo" phrasing - falling back to that construction directly if the builder returns nothing. There is no model, no image analysis, and no network request: the tool has no access to the image files at all and is combining strings you typed.