Every creator knows the feeling. You open the search to find one reference, and an hour later you have forty tabs, three saved collections, and no image. The search was supposed to take five minutes. The search took the afternoon. The work never started, because the hunt never ended.
Searching feels like progress. It is not. Searching is preparation, and preparation has a finish line. This guide is about that finish line: how to know when you have enough, how to stop without feeling like you are giving up, and how to turn the pile of references into the first generation. The goal is not to search less. The goal is to search until the hunt is done, and then to create.
クイック判定
Stop when you can describe the image in one sentence without opening another tab. Three references are enough for most images — one for the pose, one for the light, one for the mood — and every tab after the third is procrastination wearing a research costume. Set a ten-minute timer, close the search when it rings, and write the brief from memory: whatever you remember is what actually mattered, and the hunt is done when the brief writes itself.

このガイドでは
- 1. The search that never ends
- 2. Three signs the hunt is done
- 3. The ten-minute rule
- 4. From a pile of references to one brief
- 5. The search budget: time, tabs, and taste
- 6. When more searching actually helps
- 7. The stop-and-start checklist before you generate
- 費用: 検索は無料だ, re-rolling is not
- 8. 同意: すべてを決めるルール
- FAQ
Searching vs creating
| Searching | Creating | |
|---|---|---|
| Feels like | Progress, because the pile grows | Risk, because the first result might be bad |
| Costs | Time only | Time and credits |
| Produces | References you may never use | Images you can keep, fix, or throw away |
| Ends when | Never, unless you decide it ends | When the image exists |
| Failure mode | Forty tabs, no image | One bad image, one lesson learned |
| What it teaches | What exists | What you actually want |
The table is the whole argument. Searching teaches you what exists; creating teaches you what you want. You cannot learn what you want from a search results page, because the search page never asks you to choose. The first generation forces the choice, and the choice is the work.
1. The search that never ends
The search never ends on its own because search engines are built to never end. Every result page has another page behind it, every image has a similar-images link, every tag has a related tag. The system is designed to keep you looking, and it is very good at its job. If you wait for the search to run out of results, you will wait forever.
The hunt also never ends because of a quiet fear: the fear that the perfect reference is one more search away. It is a reasonable fear and a useless one. The perfect reference does not exist, because a reference is not the image — it is a starting point. A good reference is one that answers a specific question: what pose, what light, what mood. Once the question is answered, the reference has done its job, and the next search is not research, it is delay.
2. Three signs the hunt is done
The first sign is the sentence test. Close the search, look away from the screen, and describe the image you want in one sentence. “A woman in a teal dress closing a laptop in a warm living room, evening light, relieved smile.” If you can say it without opening another tab, you have enough. If you cannot say it, you do not need more references — you need five minutes of thinking, and thinking does not happen in a search box.
The second sign is the three-reference rule. One reference for the pose, one for the light, one for the mood. Three is enough for most images, and five is the absolute ceiling. If you have eight references, you do not have a clearer brief — you have eight competing briefs, and the generation will try to satisfy all of them and satisfy none of them. Pick three, delete the rest, and the brief gets sharper instead of vaguer.

The third sign is the repetition test. Search the same idea with three different wordings. If the same three or four images keep appearing, you have mapped the territory. There is nothing left to find that would change the image. The hunt is done when the search stops surprising you — and the moment it stops surprising you is the moment to close it.
3. The ten-minute rule
Set a timer for ten minutes before you open the search. Ten minutes is long enough to find three references and short enough that the search cannot eat the afternoon. When the timer rings, close the search whether you feel ready or not. The feeling of unreadiness is not information; it is the fear from section one, wearing a different coat.
The timer works because it moves the decision from your feelings to the clock. Without a timer, the decision to stop is a negotiation with yourself, and you will lose that negotiation every time — there is always one more search that feels necessary. With a timer, the decision is made in advance, and all you have to do is obey it. Obeying a timer you set yourself is not giving up; it is keeping a promise.
4. From a pile of references to one brief
A pile of references is not a brief. A brief is one paragraph that says what the image is, and the way to write it is to close the references and write from memory. Whatever you remember is what actually mattered. The details you forget were never load-bearing; they were decoration, and the generation does not need them.
Write the brief in three lines. Line one: the subject and the pose. Line two: the light and the room. Line three: the mood and the one detail that makes it yours. “A woman in her late twenties in a teal dress, closing a laptop on a coffee table. Warm living room, evening lamps, soft shadows. Relieved smile, the feeling of work finished.” That is a brief. It fits in a prompt box, and it leaves no room for the generator to guess what you meant.
If the brief comes out longer than three lines, cut it. A long brief is usually three briefs arguing with each other, and the generator will pick the loudest one. Cut until only one image remains in the paragraph. Then generate from that paragraph, not from the pile. The pile stays closed. The pile has done its job.
5. The search budget: time, tabs, and taste
Give the search a budget the way you give a generation a budget. Time: ten minutes. Tabs: three references kept, the rest closed. And taste: one decision about what you will not do. The taste decision matters most, because it is the one the search cannot make for you. “No studio backgrounds” or “no looking at the camera” or “warm light only” — one rule, written down before the search starts, and every reference that breaks it gets closed on sight.
The budget also protects your taste. Taste is not a mysterious gift; it is a set of decisions repeated until they become instinct. Every hour spent scrolling trains the opposite instinct: the instinct to accept whatever the search shows you. A budget forces you to choose, and choosing is how taste is built. Ten minutes of choosing teaches more than an hour of scrolling.

6. When more searching actually helps
More searching helps in exactly two cases. The first is when you genuinely do not know what the thing looks like — a period costume, a specific animal, a place you have never seen. Then the search is not delay, it is research, and the timer should be longer: thirty minutes, with a note at the end of what you learned. The difference is that research ends with a note. Delay ends with forty tabs and no note.
The second case is when the first generation fails in a specific, nameable way. “The pose is wrong because the reference pose was wrong” is a searching problem, and one targeted search fixes it. “The pose is wrong and I do not know why” is a thinking problem, and no amount of searching fixes it. Learn to tell the two apart: if you can name what is missing, search for it. If you cannot name it, close the search and think.
Everything else is delay, and delay has a tell: the searches get vaguer as they go on. A research search gets more specific with every query. A delay search gets broader — from “woman closing laptop warm light” to “cozy aesthetic” to “evening vibes” — because the searcher is no longer looking for anything, only looking. When the words get vaguer, stop.
7. The stop-and-start checklist before you generate
Run this list before the first generation. Timer set and rung, or ten minutes spent — the search is closed either way. Three references kept, the rest closed. One sentence describing the image, spoken without looking at the screen. A three-line brief: subject and pose, light and room, mood and one detail. One taste rule written down. If any item is missing, fix it now, before you spend a credit. The checklist takes two minutes. The afternoon it saves is the whole point.
The hardest item is the last one to check: is the search actually closed? Not minimised, not in another tab, closed. A search left open is a door left open, and you will walk back through it the moment the first generation disappoints you. Close it. The references are saved. The brief is written. The next step is a generation, 検索ではない, and the generation is the work you sat down to do.
費用: 検索は無料だ, re-rolling is not
Searching costs nothing on this site. Text-based searching is completely free with no registration, クレジットカード不要, and no account required — free users get unlimited text-based searches. The search is the cheapest part of the whole workflow, which is exactly why it is so easy to overuse: free things feel harmless, and harmless things expand to fill the afternoon.
What costs money is the generation that follows, and the re-rolls after it. A standard text-to-image generation is 6 クレジット; an image-to-image or template generation is 49 クレジット. A vague brief — the kind written from forty tabs instead of three references — needs three or four re-rolls to land, and each re-roll is another 6 or 49 クレジット. A sharp brief, written from a closed search, often lands on the first try. 価格確認済み 13 11月 2026. Pack prices for reference: $19.99 〜のために 1,000 credits and $59.99 〜のために 5,000 クレジット, both one-off payments with credits added immediately; annual packs are $119.88 と $359.88, paid once and issued monthly over twelve months. At the $19.99 pack a credit works out at about $0.020, and at the $59.99 pack about $0.012.
Do the arithmetic once and the ten-minute rule stops feeling strict. Four re-rolls at 49 クレジットは 196 クレジット——およそ $3.92 at the $19.99 pack rate — spent because the brief was written from a pile instead of a paragraph. Ten minutes of searching, three references, one brief: that is the cheapest generation you will ever make.
8. 同意: すべてを決めるルール
Generate only images of yourself or of fictional adult characters you invented. Never generate an image of a real person without their consent, and never use a real person’s photo as the input for a transformation they did not agree to. The rule covers the search too: do not search for real people’s photos to use as generation inputs. A reference found in a search is not permission, and a photo that exists online is not a photo you may transform.
The consent rule is also what makes the workflow in this guide work. Fictional characters need no permission, so the search can be fast and the brief can be bold. You can describe exactly the character you imagined — the dress, the room, the light, the mood — without borrowing anyone’s face. The ten-minute rule works because the character is yours. If the character were a real person, no amount of searching would make the generation acceptable.
FAQ
How many references do I actually need?
Three: one for the pose, one for the light, one for the mood. Five is the absolute ceiling. More than five and the references start competing with each other, and the generation tries to satisfy all of them and satisfies none.
What if ten minutes is not enough to find what I need?
Then the search is too vague, not too short. A specific search — subject, pose, light, room — returns usable references in minutes. If ten minutes is not enough, rewrite the search to be more specific instead of giving it more time.
How do I know the hunt is done?
Three signs: you can describe the image in one sentence without opening another tab; you have three references and the fourth would only repeat them; and the same images keep appearing under different search words. Any one of them means stop.
Should I save references for later?
Save the searches, not the images. Keep a note of the exact words that returned the pose, the light, and the mood you wanted. After a month you will have a personal search vocabulary, and ten minutes will start feeling generous.
Does searching cost credits?
いいえ. Text-based searching is completely free with no registration, クレジットカード不要, アカウント登録不要. What costs credits is the generation — 6 credits for a standard image, 49 credits for image-to-image — and the re-rolls that a vague brief makes necessary.
What if the first generation is bad even after a good search?
Then the problem is the brief, not the search. Read the brief out loud: if it describes two different images, cut it to one. If it describes one image clearly, the generation was a miss, not a mistake — re-roll once from the same brief before you change anything.
続きを読む
- What Your Search History Says About Your Taste
- Searching for Specific Kinks Without Doomscrolling
- Search by Mood, Not by Keyword
Start the search in the AI porn search — text-based searching is free and unlimited — and when the hunt is done, turn the brief into an image with the AI porn generator.