Help for ANIMANIQ

Every world has its own rules: what may and may not be drawn in it. The world’s generator sets them, and they are named here as it names them:

No world allows children or teenagers in any explicit form. An image with forbidden content is rejected by the check, and its tokens are returned.

Tokens for an image are charged when the picture is ready. If it fails, they are returned in full.

“Improve prompt” is charged for every press. You can bring your own text back while you stay on the page — the tokens for improving are not returned.

The face button above the prompt field opens your cut-outs. Tap a cut-out to tick it for generation; the ticked ones are shown next to the button.

You can name a face right in the prompt: <face:1> — the first of the faces ticked in the faces window, <face:Name> — a person’s face from the faces window by name, even if nothing is ticked. If the prompt has such tags, only the faces they name go into the image; the generator does not see the tags themselves.

You can set the frame ratio with a tag, like in Midjourney: <ar:5:7> or <ratio:5:7> — numbers from 1 to 99, from 1:3 to 3:1. The tag overrides the ratio button; pressing the button rewrites the tag. A ratio out of range turns red and the job is not sent.

Whether the generator accepts cut-outs right now is said in the cut-outs window itself. While it does not, the image comes out without the face swap.

An image with a cut-out cannot be published in the gallery: it stays in your private gallery. This was decided in advance so that someone else’s face does not end up on display.

Flux (tag <flux>) understands Russian. The generators <uv>, <sd> and <jug> read only English: they will not read Russian words, and the picture will miss them.

The “Improve prompt” button translates a prompt in any language — Russian, French, German and others — into English: the translation is shown in the field and is free. If the prompt still has words not in English and the generator does not read them, the bar at the bottom of the window says so in red when you press “Create” — once per prompt. Examples of English prompts are in the section “How to write a prompt”: the arrow next to an example puts it into the field.

The button rewrites your prompt the way the generator reads it best: for Flux — as connected paragraphs, for the others — as short English phrases. Tags in angle brackets stay as written.

The button says what the next press will do. The cycle of steps depends on who draws. For Flux there are three steps: “Improve · + details and setting” (objects of the place and textures), “Improve · + frame work” (angle, shot, light and shadow), “Improve · new scene” (another scene with your hero); the prompt can be written in any language. For SD models (UV and others) the first step is “Improve · formatting” — the text becomes short English parts separated by commas, nothing new is added — and a prompt not in English is first translated for free (“Improve · translation”). A new scene keeps your hero — the first part of the prompt, where the person is named — and changes the place, pose, light and frame around them. Add something of your own and the cycle starts again.

The price is in the “Prices” line above.

Models used to work with a list: girl, roof, sunset, wind, detailed. Our generator does not read it that way. It understands an ordinary sentence: who is in the frame, what they are doing, where it happens.

A young woman in a school uniform stands on a roof and watches the sunset, the wind lifts her hair.

It will accept a list of words too — but it will assemble the picture from them at random, because it will not understand what goes with what.

You set the weight of words by their order. The generator does not understand brackets, asterisks or multipliers — it reads them as ordinary characters. Start with what must be in the frame for sure and move secondary things to the end.

The order that works: who → what they are doing → where and in what light → how it is shot.

Every word splits attention. In a prompt of forty words each one gets little weight, and the picture comes out “a bit of everything”. Two or three clear sentences work better than a paragraph.

If the result is “almost it”, remove the extra rather than adding more.

The words “beautiful”, “high quality”, “masterpiece”, “highly detailed” add nothing: they are words about the picture, not about what is in it. But where the light falls from and from what distance it is shot change the whole frame.

Useful words: backlight, soft diffused light, close-up, wide shot, low angle.

Without it the generator picks a medium shot by itself. Say it — and you get what you wanted: a face filling the frame, the whole figure, a scene from above.

And there will be none while this model is in use. It is built so that it does not compute “what not to draw” at all — such a field would be a button that does nothing. So write what MUST be in the frame, not what must not.

“Without a frame” inside a prompt reads as “frame”: the generator does not know the word “without”.

We replace a name from the list with the key word of its style set and plug the set into the generation. If that word is missing from the final text, the set loads and does not work — the picture comes out without the character. So a job with a lost key does not start at all, and the tokens are returned.

The highlighting in the field shows what was recognised: names in one colour, techniques in another, places in a third, clans in a fourth.

It replaces character names with set keys and puts the key as the first word — that way the set weighs the most. It adds the hero’s look, and for those without a trained set also their distinctive features: there the text is the only carrier of the character. It puts the world tag last. Everything else goes letter for letter — not a word is added or removed.

It rewrites the scene around the hero: place, light, weather, angle. The price is in the line “Prices”. It does not touch your words or your character — only what is around them changes.

Every fifth press goes further and changes the kind of picture itself: a portrait becomes wide action, calm becomes catastrophe, day becomes night. The character stays themselves there too. The bar at the bottom of the window says what the press was.

If the model returned almost the same thing, we do not count it as an improvement: the site puts in a scene so that the picture really changes.

Spoiled the prompt? The undo button gives back your original text in full — it is kept until you leave the page. Tokens are not returned on undo.

Many people improve at once, so there is a queue. If we say “too many improvements right now”, no tokens were charged — press again in a minute.

This is not a mistake in your prompt — that is how the fast model works. But you can influence it, and what influences it most is the pose.

A standing figure comes out cleaner than a sitting one. One person cleaner than two. Two people who touch, hug or hand something to each other are the hardest case: that is exactly where extra limbs appear.

If the hands came out crooked, first simplify the pose, rather than adding words. “Stands and looks at the camera” fixes more than ten clarifications about fingers.

Asking for “correct hands”, “five fingers”, “anatomically correct” is useless: the model does not understand such demands as a ban, and often draws the hands bigger and more noticeable.

In order of frequency: a list of tags instead of a sentence; too many words; the description of appearance argues with the character’s style set; a contradiction inside the prompt (“night” and “sunlight”); neither light nor angle named.

The cost is given in the “Prices” line. Tokens are not taken when you press, but when the job goes into assembly — that is, when it actually starts being computed. They are charged when the picture is ready.

There is no refusal. The job joins the queue and waits, just as it waits for a returning service. Top up your balance and it goes into assembly by itself, without a single press.

Yes, as many as you like. There is no limit on how often you press: each press is paid and becomes a separate job in the queue. One is computed at a time, the others wait their turn — but are not lost.

They are returned in full. No exceptions: no picture — no payment. The card keeps the reason for the failure.

Puts it back at the end of the queue. It is not an attempt to get the same picture: the retry has its own seed, so the result will differ — and that is right, because there was no picture to repeat in the first place.

For the same reason, pressing again with the same prompt will not give the same picture: the seed is new every time.

The prompt or the finished image did not pass the check. Tokens are returned. Change the prompt — what most often gets in the way in this world is described under “How to write a prompt”.

The generator draws one image at a time. Yours waits its turn; its place is shown on the card. No need to keep the tab open — the job continues.

The site talks to you through one line at the bottom — your balance, your place in the queue, the reason for a refusal. It appears when there is something to say and slides away by itself. A normal message is pale, yellow means “not everything works”, red means “not working”.

Ctrl+Alt+D builds a diagnostics report: a screenshot, the event log and the passports of your generations. Send it to support if something goes wrong.

The site build you are looking at: —. Mention it when reporting a problem — the number shows exactly what you have open.

The payment window shows only the wallet address and the exact ETH amount. The amount is unique — that is how the payment is recognised, so send exactly that.

The image goes to the public gallery and to your author page. Everyone will see it, including people who are not signed in.

The public gallery shows the newest images first, without curation: the later an image is published, the higher it stands.

A separate page with all your published images. You can share its link with anyone — no sign-in is needed to view it. The cover, the avatar and the name are changed in your profile: click the avatar or the name right there. The “Open my author page” link is in your profile, to the right of the name.

The country flag appears next to your name on every image. It is your choice: if you do not want the badge, do not set a country.

Yes, with the same action on the card. The image comes back to you and disappears from the public gallery.