Every volume here has a number, a shelf, and a way back to it. Your prompts deserve the same.
Every volume here has a number, a shelf, and a way back to it. Your prompts deserve the same.
Modular blocks, a render log, and the end of retyping the same prompt forever
Open your prompt box right now and look at what is in it. If it is a single wall of comma-separated words you typed forty minutes ago and could not reconstruct tomorrow, you are not alone, and you are quietly throwing away most of your own work. Every artist I know has the same graveyard: a folder of images they love and no idea how they made them.
The fix is not a better model or a longer prompt. It is a filing system. Painters keep swatch books. Photographers keep their lighting diagrams. Somewhere along the way AI art convinced everyone that the prompt is disposable, a thing you improvise and discard, and the result is thousands of people rediscovering the same lighting phrase every three weeks.
Here is the mental shift that makes everything else possible. A prompt is not a description. It is a stack of independent modules, and each module answers a different question:
Once you can see those seams, a prompt stops being a sentence you compose and becomes a recipe you assemble. Swap the subject block and keep everything else, and you get the same character in the same world doing something new. Swap the light block alone and you get the same shot at golden hour. That is a library.
Keep each one short enough to paste and specific enough to be worth saving. A light block might be: late afternoon window light from camera left, soft falloff, warm highlights, gentle shadow on the far cheek. A camera block might be: 85mm portrait lens, f/1.8, shallow depth of field, chest-up framing, eye level. Neither one mentions a person. That is the point. They are interchangeable parts.
Give every block a name you would actually use at midnight. "Window Soft Warm" beats "Lighting Preset 4" forever. You are writing for the version of yourself who is tired and impatient and just wants the good one.
A prompt on its own is half a record. The other half is the settings that produced the image, and this is where almost everyone loses their best work. Store these next to every prompt you keep:
Prompts are not portable across models the way people assume. A phrase that produces beautiful rim light on one checkpoint can do almost nothing on another, and if you did not write down which model made the image, your library slowly fills with entries that no longer work and you have no way to tell which ones. The model name is the single most valuable field in the whole system.
Boring is correct here. A spreadsheet with a row per saved render and a column per field will outperform any clever tool, because you will still be able to read it in two years. Name your image files so they carry their own identity, something like 2026-07-21_portrait_windowsoft_003.png, and put that filename in the row. Now the picture and the recipe can always find each other.
Keep a second, much shorter sheet for the blocks themselves, one row per block, grouped by category. That is the part you will paste from daily. The render log is your history. The block sheet is your working palette. Do not merge them.
Save on delight, not on completion. The instinct is to log everything, which turns into a chore, which turns into abandoning the whole system by Thursday. Instead: the moment an image makes you sit up, stop and record it before you generate anything else. Ten seconds, right then. Everything else can go unlogged forever and the library will still be the most valuable folder you own.
The honest counterpoint is worth saying out loud, because systems like this can absolutely go wrong. If you build the library too early, you will lock in your first tolerable version of everything and stop exploring. A library is a floor, not a ceiling. Some of the best images you will ever make will come from ignoring the whole thing and typing something strange at one in the morning. Keep that door open. Log the result if it works.
Three things, and they arrive in this order. First, your setup time collapses, because you are assembling instead of inventing. Second, your work starts to look like it came from one person, which is the actual definition of style and something scattered prompting almost never produces. Third, and best, you start improving deliberately: when a render disappoints, you can change exactly one block, regenerate, and learn something, instead of rewriting the whole prompt and learning nothing.
That last one is the real prize. A library does not just save you typing. It turns every session into an experiment with one variable, and that is how anyone gets good at anything.
If you want the raw material to start filling those blocks, pull your light vocabulary from the lighting guide, your camera block from the lens language guide, and your wardrobe block from the fabric and clothing texture guide. Copy them into your sheet, name them, and you have a working library before you generate a single image.