Gidsen

The Keeper Rate: Measuring Whether a Generator Is Actually Good for You (2026)

Every generator promises the same thing: great images, fast, cheap. Every gallery looks the same too: twenty perfect samples, all keepers. The gallery is marketing. Your own history is the truth. Over the last month, how many of your generations did you actually keep? How many did you delete, re-roll, or quietly abandon? That ratio — keepers divided by generations — is the only honest score a generator can get, because it measures what the tool does for you, not what it does for its own advertisements. A generator that gives you 30 keepers out of 100 generations is doing twice the work of one that gives you 15, at the same price per generation. That ratio has a name in this guide: the keeper rate.

Snel oordeel: Track your next 100 generations and count the keepers. A keeper is any image you save, use, or post — not one you almost liked. Divide keepers by 100 and you have your keeper rate. Compare that rate across tools, modes, and months, and judge every generator by it, because the keeper rate is the verdict on the tool, not the gallery. Then convert the rate into money: at 6 credits per standard image, a 25% keeper rate means each keeper costs 24 credits, or $0.48 op de $19.99 pakket. Een 10% rate means each keeper costs 60 credits, or $1.20. Same tool, same price, two very different costs — and the difference is yours to measure.

A glamorous woman in a dark green evening gown beside the title text The Keeper Rate — Score Your Generator

Contents:

  1. What the keeper rate is, and what it is not
  2. How to measure it: the 100-generation log
  3. Reading your rate by mode
  4. From rate to money: the cost of one keeper
  5. What drags your rate down (it is rarely the tool alone)
  6. Raising your rate without spending more
  7. What a keeper costs, in credits and dollars
  8. What the keeper rate cannot tell you
  9. Consent and the rules
  10. FAQ
Modus What counts as a keeper What to log for each generation What a falling rate usually means
Text to image Saved, used, or posted Prompt, seed, keep or delete Prompt drift or a changed model
Image to image Usable without a second pass Source photo, strength, keep or delete Weak reference photo or wrong strength
Gezichtswissel Face fits the photo at a glance Source face, target photo, keep or delete Light or angle mismatch
Video template A clip you would show someone First frame, length, keep or delete Bad first frame or too much motion asked

1. What the keeper rate is, and what it is not

The keeper rate is a fraction: the number of generations you keep divided by the number you make. Make 100 standaardafbeeldingen, keep 25, and your rate is 25%. Make 40 image-to-image passes, keep 8, and your rate is 20%. That is all it is, and that simplicity is the point. It does not care about resolution, model names, or what the gallery promised. It cares about one thing: out of everything you paid for, how much survived.

It is not a quality score, and it is not a review. Een 40% rate does not mean the tool is good in any absolute sense — it means the tool is good for you, with your prompts, your reference photos, and your taste. The same generator can give one person a 35% rate and another a 12% rate, and both numbers are correct, because the second person is asking for harder things: two figures, unusual angles, a face that has to survive ten images in a series. The rate is personal, and that is why it beats every ranking list: the list measures the tool, and the rate measures the tool plus you.

2. How to measure it: the 100-generation log

You need a notebook or a spreadsheet, and one rule: every generation gets a row ending in one of four words. Keeper: you saved it, used it, or posted it. Fixable: one edit or re-roll could save it. Miss: the idea was right and the execution was wrong. Broken: hands, anatomy, or a face that fell apart. Four buckets, no maybes — the moment you allow “almost a keeper”, the log stops meaning anything.

Log the next 100 generations in one mode before you judge anything. Log the prompt or the source photo, the seed if you set one, and the verdict. A row takes ten seconds, so the whole log costs under twenty minutes of bookkeeping for a month of evidence.

3. Reading your rate by mode

Text to image is the baseline. You write a prompt, the tool draws a still, and you keep it or you do not. A rate in the twenties is a normal working rate for a careful prompter: one keeper in four or five generations. Below 10%, something is off — usually the prompt is trying to do too much at once, or the same seed is being re-rolled so often that the variety is gone. Above 40%, you are probably being too easy on yourself, keeping images you will never use, and the log is flattering you.

Image to image runs lower, and it should. The tool has two masters: your prompt and your source photo. When they disagree — the photo says one pose, the prompt says another — the result satisfies neither, and the rate falls. The fix is to pick a side before you generate: either the photo leads and the prompt describes what is already there, or the prompt leads and the photo is only a loose reference at low strength. The strength slider is the argument between the two, and most people set it once and never touch it again.

A woman reviewing printed photos spread across a wooden desk under a warm lamp

4. From rate to money: the cost of one keeper

The keeper rate turns into money with one division: credits per generation divided by the keeper rate. A standard image costs 6 credits. At a 25% rate, each keeper costs 24 credits: $0.48 op de $19.99 pakket, $0.29 op de $59.99 pakket. At a 10% rate, each keeper costs 60 credits: $1.20 or $0.72. The generation price never changed. The keeper did.

Run the same division for the other modes, and the ranking of “cheap” en “expensive” changes completely. An image-to-image pass costs 49 credits. At a 20% keeper rate, each keeper costs 245 credits: $4.90 op de $19.99 pakket, $2.94 op de $59.99 pakket. A face swap costs 26 credits. At a 50% rate, each keeper costs 52 credits: $1.04 or $0.62. A five-second video template costs 351 credits. At one keeper in three clips, each keeper costs 1,053 credits: $21.06 or $12.64.

Those numbers are the real price of your work, and they are the numbers to compare when someone asks whether a generator is cheap. A tool that charges 6 credits per image and gives you a 10% rate is more expensive per keeper than a tool that charges 8 credits and gives you a 25% rate. The per-generation price is advertising. The per-keeper price is the bill. And the per-keeper price is the number that decides whether a pack lasts you a week or a month: 1,000 credits at a 25% still rate buys about 41 bewaarde afbeeldingen; at a 10% rate it buys about 16.

There is one more comparison the rate makes possible, and it is the one that saves the most money: the same prompt on two generators. Run your five best prompts on a second generator, log the keepers, and compare the per-keeper cost, not the per-generation price.

5. What drags your rate down (it is rarely the tool alone)

The most common drag is the overloaded prompt. A prompt that asks for a pose, a setting, a light direction, a mood, a camera lens, and a style in one sentence asks the model to solve six problems at once, and it will solve three well and fudge the rest. The fix is to split the work: let the first generation settle the composition, and let a second pass — an image-to-image edit at 49 credits — add the detail. Two generations with a plan beat five generations of hoping.

The second drag is the reference photo. For image-to-image and face swap, the photo is half the prompt, and a bad photo caps the rate before the tool even starts. Blurry, dark, side-lit, or cluttered photos produce blurry, dark, side-lit, or cluttered results, and no prompt rescues them. The log shows this as a broken bucket full of images that all share one thing: the same weak source. Pick a sharper photo, crop tighter on the face, and match the light. A ten-minute photo cull is worth more than a hundred credits of re-rolls.

A woman in a cosy armchair comparing two coastal photos on her phone

6. Raising your rate without spending more

The cheapest way to raise your rate is a test batch. Before a real session, run five standard images on the prompt you are about to use heavily, at 6 credits each — 30 credits, $0.60 op de $19.99 pakket, $0.36 op de $59.99 pakket. If three of the five are keepers, the prompt is ready, and the session will be efficient.

The second way is to keep a keeper shelf: your ten best outputs, prompts and settings attached. Re-run one whenever you try a new tool or a new month. If it still produces a keeper, your skill is intact and the tool is the variable. If it does not, something changed upstream, and you will know it in one generation instead of fifty.

The third way is to stop at the keeper. Most people keep generating after the keeper arrives, because the next roll might be better — and the rate counts every roll after the keeper as a miss. Set a rule before the session: one keeper per idea, then move on. The discipline feels like settling, but it is the difference between a 25% rate and a 12% rate on the same prompts, with the same tool, on the same day.

What a keeper costs, in credits and dollars

Put the whole method on one ledger. One hundred standard images at 6 credits each is 600 credits: $12.00 op de $19.99 pakket, $7.20 op de $59.99 pakket. At a 25% keeper rate that buys 25 keepers at 24 credits each, $0.48 or $0.29; at a 10% rate it buys 10 keepers at 60 credits each, $1.20 or $0.72. One hundred image-to-image passes at 49 credits is 4,900 credits: $98.00 or $58.80. At a 20% rate that is 20 keepers at 245 credits each, $4.90 or $2.94. Three five-second video clips at 351 credits is 1,053 credits: $21.06 or $12.64 for one keeper at a one-in-three rate. Prijzen gecontroleerd 13 nov. 2026. Pack prices are $19.99 voor 1,000 credits and $59.99 voor 5,000, both one-time purchases with credits arriving immediately; annual packs exist at $119.88 en $359.88, paid once with credits issued monthly across twelve months. There is no automatic renewal on any of these.

What the keeper rate cannot tell you

The keeper rate cannot tell you whether an image is good. It can only tell you whether you kept it, and you are a biased judge of your own work. A rate can be high because your taste is undemanding, or because you keep everything out of guilt for the credits spent. The log measures discipline and fit, not beauty, and it should never be quoted as a quality score to anyone else.

It also cannot compare two people. Your 25% and someone else’s 40% do not mean their tool is better — they may be asking for easier things, or keeping more generously. Compare your rate with your own past rate, on the same mode, with the same strictness about what counts as a keeper. That comparison is honest.

A keeper rate does not change the rules, and a high rate on a forbidden use is still a forbidden use. Generate only images of yourself, or of fictional adults you created. Never upload a photo of a real person to undress them, swap their face, or place them in a scene they did not agree to — and do not use a high keeper rate as proof that a use is fine: a tool that does something well is not a tool that does something right. Keep your reference photos yours, keep your characters fictional adults, and remember that the images you keep are the images you are responsible for, long after the credits are spent.

FAQ

What is a good keeper rate for text-to-image generations?

For most careful prompters, a working rate sits between 15% en 35% — one keeper in three to seven generations. Below 10%, the prompt is usually overloaded or the idea is unfinished. Above 40%, you are probably keeping images you will never use. Compare your rate with your own past rate, not with anyone else’s.

How many generations do I need to log before the rate means anything?

One hundred in a single mode. At 20 generations, one lucky streak moves the rate by five points. At 100, the number is stable enough to act on. Log the prompt or source photo, the seed, and one of four verdicts: keeper, fixable, miss, or broken.

Why is my video keeper rate so much lower than my image rate?

Because a clip has to hold together across every frame, and one bad frame ruins it. One keeper in three clips is a working video rate. Price it per keeper: three five-second clips at 351 credits is 1,053 credits, $21.06 op de $19.99 pakket, for one keeper.

How do I work out the cost of one keeper?

Divide the credits per generation by your keeper rate. A 6-credit standard image at a 25% rate costs 24 credits per keeper: $0.48 op de $19.99 pakket, $0.29 op de $59.99 pakket. At a 10% rate it costs 60 credits: $1.20 or $0.72. The per-generation price is advertising; the per-keeper price is the bill.

Can the keeper rate tell me whether to switch generators?

Ja, if you run the same five prompts on both tools and compare per-keeper cost. A tool that charges twice as much per generation but gives three times the keeper rate is the cheaper tool. Do not compare gallery samples — compare your own log.

What is the fastest way to raise my keeper rate?

Run a five-image test batch before a real session: 30 credits, $0.60 op de $19.99 pakket. If three of five are keepers, the prompt is ready. If none are, fix the prompt before spending more. And stop at the keeper — one keeper per idea, then move on.

Lees verder: When Your Generator Changes: Price Hikes, Model Swaps & Vanishing Features · Switching AI Porn Generators: What Moves With You and What Starts Over · What $20 Buys: A Month of AI Porn on 1,000 Credits

Tools on this site: AI-pornogenerator · AI-pornovideogenerator · AI-uitkleden · AI Afbeelding Editor

Inloggen

Registreren

Wachtwoord opnieuw instellen

Voer je gebruikersnaam of e-mailadres in, je ontvangt een link per e-mail om een nieuw wachtwoord aan te maken.