How we rate
This page exists so that anyone can redo the calculation. There is no “exclusive ranking” and no secret criterion: seven criteria, fixed weights, and the sum is the score.
The seven criteria and their weights
| Criterion | Weight | How it scores |
|---|---|---|
| Pricing model | 3.0 | Free 3.0 · Freemium 2.0 · Free trial 0.8 · Paid 0.2 · Price not public 0 |
| Output types | 1.75 | 0.5 per type among image, video, chat and voice, up to 1.75 |
| Filtering | 1.5 | Uncensored 1.5 · Partial 0.75 · Strict 0.15 |
| Testing and freshness | 1.5 | Quota measured in our test 0.75 or stated limit 0.2, plus 0.75 if the verification is less than 180 days old |
| Spanish interface | 1.0 | Interface in Spanish 1.0 · Partial 0.5 · Not verified 0.5 (neutral) · English only 0 |
| Sign-up | 0.75 | 0.75 if no account is required, 0 if one is |
| Chat | 0.5 | 0.5 if the tool holds a conversation, 0 if it only generates image or video. On the Image generator and Video generator pages that half point is not counted, because nobody there is looking for conversation |
Maximum total: 10. The score is recalculated automatically every time a tool card is saved. Nobody types the score by hand.
The weights changed on 21 September 2026, and we say so here because changing a scale without notice would be the best way not to deserve trust. Price went up from 2.5 to 3 points, output types from 1.5 to 1.75, and mandatory sign-up went down from 1.5 to 0.75. The reason: requiring an account is the industry norm and says little about quality, while at 1.5 points that binary criterion sank every paid tool at once and left the ceiling of the ranking at 7.2 out of 10. With the new weights, the top tool is at 8.3 and there is still no 10.
The three levels of testing
Each row in the ranking says where the information about the free tier and its limits comes from. There are three situations, and we do not mix them:
- Price not public. We did not find the figure before signing up. We note what the page states, with the date, and the tool gets zero on the price criterion, which is worth 2.5 points. It stays in the ranking, with the label “Price not public”.
- Stated by the vendor. We read the vendor’s own pricing page or subscription terms and copy what is there, with the date. That is most of the list today.
- Measured in our test. We open the tool with an account and note what it actually delivered. It weighs more because it costs more and because it sometimes contradicts what the site states.
A stated limit is never presented as measured. If you see “What the site says” in the row, that is the vendor’s word; if you see “What we measured”, it is ours.
What the four price labels mean
Free: usable without a card and without a mandatory account, and the quota does not block normal use. Freemium: requires an account, has a free quota that can be measured, asks for no card to start. Free trial: asks for a card up front, or gives a one-off quota that does not renew. Paid: nothing usable without paying.
When a vendor hides the pricing table behind sign-up, the label becomes “Price not public” and the tool loses the points for that criterion. Hiding the price is a decision, and the score reflects it.
Who runs the tests
The measurements are made by Lia, the AI that signs the site: she opens the tool, counts what the free tier delivers, reads the pricing table and records the date. The weights of the score, the validation of a doubtful measurement and publication are decided by the editorial team. We say so here because a calculated score is only worth as much as its method, and the method includes who presses the button.
What we do not do
- We do not sell positions. There is no sponsored slot in the ranking and no editorial ordering above the score.
- We do not test the filtering limits by producing content. The filtering classification comes from what the vendor states and from what the community reports.
- We do not fill in what we have not verified. The language was checked on 21 September 2026 by opening each site with a Spanish language header; none was left unverified: those that offer Spanish only in their settings, such as JanitorAI, Muah.AI or OurDream, get half a point rather than the full point. The language checked is Spanish, the ranking’s original language.
- We do not recommend any tool for use with images of real people without their consent, nor for any content involving minors.
Affiliate links
Some links in the ranking are affiliate links: if you subscribe through them, the site receives a commission. When that is the case, “Affiliate” is written below the button. The commission does not enter any of the seven criteria and does not change the position. A tool without an affiliate programme competes on equal terms with one that has it.
Freshness and corrections
Each row shows the date of the last verification. After 180 days without a review, the tool loses the 0.75 freshness points until it is checked again. If you find an outdated price, limit or label, write to [email protected] with the link where you saw it. We correct the row and the date.