The generation finishes, the image appears, and you have about three seconds to decide what happens to it. Most people decide in those three seconds and regret it later — they keep images they should have deleted, delete images they should have fixed, and fix images that were never worth the repair. The decision is not about taste. It is about cost, and cost is something you can measure before you feel it.
Schnelles Urteil: Keep an image only if it passes a fixed standard, fix it only when the fault is small and local, and delete it when the repair costs more than a fresh run — because keep it only if you can say why it is worth keeping, und “I already paid for it” is not a reason.

Contents:
- The one-sentence version
- Keep: the standard a finished image has to meet
- Reparieren: what the editor can actually repair
- Delete: when fixing costs more than re-running
- The decision in numbers
- The 24-hour rule
- A filing habit for the keepers
- What each fate costs
- What the decision cannot do
- Consent and the rules
- Endgültiges Urteil
- FAQ
| Fate | When it applies | Kosten |
|---|---|---|
| Keep | Passes the standard, no fix needed | 0 Credits |
| Reparieren | One small, local fault | 49 Credits, editor pass |
| Delete | Fault is structural, or fix costs more than re-run | 21 credits to re-run |
| Wait | You cannot decide in three seconds | 0 Credits, 24 hours |
1. The one-sentence version
Every finished image gets one of three fates — keep, fix, or delete — and the fate is decided by two questions: does the image meet your standard, and is the cheapest way to reach the standard a repair or a fresh run? If it meets the standard, keep it. If it misses by a little, fix it. If it misses by a lot, delete it and re-run. The rest of this guide is the detail: what the standard is, what counts as a little, and how the numbers decide when your feelings cannot.
2. Keep: the standard a finished image has to meet
A keeper has to pass four checks, and all four are yes-or-no. Zuerst, the face: is it the face you wanted, clear and facing the way the prompt asked? Zweite, the body: is the anatomy plausible, with hands and feet that pass a quick look? Third, the setting: does the room, the light, and the background belong to the same picture as the person? Fourth, the feeling: would you show this image to someone whose opinion you respect, without apologising for it first? Four yes answers and the image is a keeper. Three yes answers and a small no is a fix. Two or fewer is a delete, because an image that fails half its checks is not one repair away from good.
The standard should be the same for every image in a set. If one image in a set of five passes and the other four do not, the set is not finished. Either the four get fixed or deleted, or the one gets deleted with them, because a set with one good image and four weak ones reads as four weak images with a lucky one. Write the standard down and keep it where you generate, because a standard you have to remember is a standard you will bend.

3. Reparieren: what the editor can actually repair
The editor is good at small, local repairs. A hand that is almost right. A shadow that is missing. A background detail that distracts. A colour that is slightly off. These are faults with a clear boundary — you can point at them, describe them in one sentence, and the repair does not touch anything else in the image. If the fault fits in one sentence, the editor can probably fix it. The editor is bad at structural repairs: a face that is the wrong person, a pose that does not make sense, hands with the wrong number of fingers, a body twisted in a way no body twists.
An editor pass costs 49 Credits, über $0.98 auf dem $19.99 pack and about $0.59 auf dem $59.99 Paket, and it is worth paying when the fault is local and everything else is a keeper. It is not worth paying when the fault is structural, because the repair will cost 49 credits and still leave you with an image you do not like. One repair, then stop: if the first editor pass does not fix the fault, the fault is bigger than it looked, and a second pass will cost another 49 credits for a worse result. Two editor passes cost 98 Credits, über $1.96 and about $1.18 — more than four fresh undress runs at 21 credits each.
4. Delete: when fixing costs more than re-running
Deleting feels like losing, so here is the arithmetic that makes it feel like winning. A fresh undress run costs 21 Credits, über $0.42 auf dem $19.99 pack and about $0.25 auf dem $59.99 Paket. An editor pass costs 49 Credits. That means one repair costs more than two fresh runs. If the chance that a fresh run produces a keeper is better than one in two, deleting and re-running is cheaper than repairing, and the chance is usually better than one in two, because you already know what went wrong and can change one thing in the prompt.
Delete is also the right fate when the image is fine but the idea is wrong. Sometimes the generation did exactly what you asked, and what you asked was a bad idea — the wrong pose for the photo, the wrong setting for the mood, a concept that sounded better than it looks. No repair fixes a bad idea, because the repair would have to change the idea, and changing the idea is a new generation, not an edit. And delete the duplicates: if you ran the same prompt five times and three of the results are keepers, the other two are not spares. They are clutter, and clutter is the tax on every future search.
5. The decision in numbers
Here is the decision as a table you can use without thinking. Keep costs 0 credits and is right when the image passes all four checks. Fix costs 49 credits and is right when exactly one check fails and the fault is local. Delete costs 21 credits to re-run and is right when two or more checks fail, when the fault is structural, when one editor pass has already failed, or when the idea itself is wrong. Wait costs 0 credits and is right whenever you cannot decide in three seconds.
Run the numbers on a bad session to see why they matter. Ten undress runs cost 210 Credits, über $4.20 and about $2.52. If four of the ten are keepers, three are fixes, and three are deletes, the fixes cost 147 credits and the re-runs cost 63 Credits, for a session total of 420 Credits, über $8.40 and about $5.04. If instead you delete the three fixes and re-run them, the session costs 210 plus 63 Credits, 273 in total, über $5.46 and about $3.28 — and the re-runs have a better chance of being keepers than the repairs do, because a fresh run can change the idea and a repair cannot. Feelings say the repair is cheaper because the image already exists. The numbers say the repair costs 49 and the fresh run costs 21.

6. The 24-hour rule
If you cannot decide in three seconds, do not decide. Put the image in a folder called “decide tomorrow” and look at it again in 24 hours. The rule works because the three-second decision is made by the part of you that just paid 21 credits and wants the money to have been well spent, while the 24-hour decision is made by the part of you that can see the image clearly. They are different judges, and the second one is better. The rule also protects you from the opposite mistake: deleting an image that is actually good, because you are tired, or because the last three runs were bad and the disappointment is spilling over.
A keeper deleted in frustration costs 21 credits to rebuild, and the rebuild might not be as good. Waiting costs nothing. The folder is free, the day is free, and the image will still be there tomorrow, when you can see it for what it is. Review the “decide tomorrow” folder once a week. Most images in it will have an obvious fate by then — keep, fix, or delete — and the ones that still do not are telling you something: the standard is unclear, or the image is fine and you are looking for a reason to dislike it. If an image survives a week undecided, keep it. Indecision that lasts a week is not doubt. It is the image being good enough that you cannot find a reason to delete it.
7. A filing habit for the keepers
Keepers deserve a home, because a keeper without a home becomes clutter just as fast as a failure. File every keeper the day you keep it: a folder for the subject, a name that says what it is, and the prompt that made it, kept with the image. The filing takes two minutes. Not filing takes twenty minutes next month, when you want the image back and cannot find it, and a keeper you cannot find is a keeper you have lost. File the fixes too, but separately: a fixed image should carry a note saying what was fixed, because the note tells the next version of you what the image cost and what its weak point was.
And empty the delete folder once a month. Deleted images that sit in a folder are not deleted, they are postponed, and postponed deletions are decisions you have to make twice. Delete means delete. The image is gone, the credits are spent, and the lesson — if there is one — is in the note you wrote before you deleted it, not in the file you kept just in case.
What each fate costs
Preise geprüft 13 Nov. 2026. Keep costs 0 Credits. Fix costs one editor pass at 49 Credits, über $0.98 auf dem $19.99 Paket (1000 Credits, $0.020 a credit) and about $0.59 auf dem $59.99 Paket (5000 Credits, $0.012 a credit). A second editor pass costs another 49 Credits, 98 in total, über $1.96 and about $1.18, which is why the rule is one fix and then stop. Delete costs a fresh undress run at 21 Credits, über $0.42 and about $0.25. Five fresh runs cost 105 Credits, über $2.10 and about $1.26. A test still image, for checking an idea before you commit to an undress, costs 6 Credits, über $0.12 and about $0.07.
A finished set shows the difference. Five keepers from five runs cost 105 Credits. Five runs with three fixes and two re-runs cost 294 Credits, über $5.88 and about $3.53. Five runs with three deletes re-run instead cost 168 Credits, über $3.36 and about $2.02 — a difference of 126 credits every set. Packs: $19.99 für 1000 credits and $59.99 für 5000, paid once and credited immediately; annual packs $119.88 und $359.88, paid once and issued monthly over 12 months. One free preview per IP is available.
What the decision cannot do
The decision cannot make a bad image good. Keep, fix, and delete are fates for a finished image, and a finished image that fails the standard is a finished image that fails the standard, however you file it. The decision organises the outcome. It does not improve it. Improvement comes from the next generation, with a better prompt, not from re-labelling this one.
The decision cannot replace the standard, and it cannot save a session that was never planned. If the standard is vague — “I will know it when I see it” — then every decision is a feeling, and feelings are biased by the credits already spent. And if you generate ten images without knowing what the set is for, no filing system will tell you which ones to keep, because keep means “keep for a purpose” and there is no purpose. Decide the purpose before you generate, and the fates decide themselves.
Consent and the rules
Erstelle nur Bilder von dir selbst, or of a consenting adult who has agreed to this specific use. Deleting an image does not delete the consent question behind it, and keeping an image does not create consent that was never given. Everyone in every image must be a fictional adult or a real consenting adult aged 18 or older. If the person did not agree to this use, the fate is delete, whatever the quality.
Endgültiges Urteil
Keep what passes the standard, fix what misses by a little and only once, delete what misses by a lot or costs more to repair than to re-run, and wait 24 hours whenever you cannot decide. The credits you already spent are gone, so decide on the next credits, not the last ones. A library of keepers you can find is worth more than a folder of almosts.
FAQ
When should I keep a finished image?
When it passes all four checks: the face is right, the anatomy is plausible, the setting belongs with the person, and you would show it without apologising for it first. Four yes answers is a keeper.
When is fixing worth it?
When exactly one check fails and the fault is small and local — a shadow, a hand that is almost right, a distracting detail. An editor pass costs 49 Credits. One pass only; if it fails, delete.
When should I delete instead of fixing?
When two or more checks fail, when the fault is structural, when one editor pass has already failed, or when the idea itself is wrong. A fresh undress run costs 21 Credits, less than half an editor pass.
Why does deleting feel so hard?
Because the credits are already spent, and deleting feels like wasting them. The credits are gone either way. Decide on the next credits, not the last ones.
What is the 24-hour rule?
If you cannot decide in three seconds, put the image in a “decide tomorrow” folder and look again in 24 hours. Waiting costs nothing, and the second look is made by a clearer judge.
How much does each fate cost?
Keep costs 0 Credits. Fix costs 49 credits per editor pass. Delete costs 21 credits for a fresh run. Five fresh runs cost 105 Credits, über $2.10 and about $1.26.
Weiterlesen: Old Prompt, New Model: Re-Running Last Year’s Prompts on Today’s Generator · One Image, Five Uses: Getting More From a Finished Undress · The Crop Is the Result: How Framing Changes Your Undress Before You Start