How to prompt Anima

Before your prompt, Mage adds Anima's recommended quality tags: masterpiece, best quality, score_9, score_8, score_7, year 2025, newest, highres, absurdres, very aesthetic. After your Negative Prompt, it adds a long list to avoid, among them worst quality, low quality, the low score tags, early, old, blurry, jpeg artifacts, artist name, text bubble, speech bubble, and the usual hand and anatomy mistakes.

That leaves only the rating for you to add from the usual prefix. Everything else you write should describe the picture.

Anima learned from tag lists and from written captions, so a pure tag prompt, a pure prose prompt, and a mix of the two all work. Tags give the most control over details; sentences are better at mood, action, and how things relate to each other.

In a tag prompt, keep the groups in order: rating, count tag, character, series, artist, then the general tags. Inside each group, order is free. Anima was trained with tags randomly left out, so you only need the tags that change the picture, not every tag that could apply.

In prose, write at least two descriptive sentences: a one-line prompt gives unpredictable results. A hybrid opens with the rating and count tags and carries on in sentences.

Anima knows the standard Danbooru tags for the camera and the light. One framing tag, a gaze or pose, and one or two light tags make the picture read like a composed shot.

Anima knows anime characters released up to about September 2025. Name the character with their series tag, then describe what they look like every time: hair, eyes, outfit. This matters most with two or more characters, together with the count tag (2girls, 1boy 1girl) and a separate description for each.

Saved Characters, References, and uploaded images do not work with Anima. To keep an original character the same across a set, reuse the same character tags word for word, keep the Seed fixed, and change only the tags that set the scene.

With the Prompt Weighting setting on, which is the default, parentheses set weights: (tag:1.8) makes a tag count for more, (tag:0.6) for less. Anima needs much stronger weights than older anime models. Use 1.5 to 2 for a visible change, and weight only one or two tags.

Every parenthesis that is part of a tag must be escaped as \( and \), or Anima reads it as a weight. That covers series qualifiers and style tags such as 1990s \(style\) and watercolor \(medium\).

Anima's own syntax marks an artist with @, but on Mage @ marks a saved Character, so a prompt with @ and a name stops before anything is generated. Write the artist's name as a plain tag in the artist slot. The effect is weaker, so back it with style tags and a sentence describing the look: line weight, shading, palette.

Anima is deliberately plain until tags give it a style. Style and meta tags such as anime screenshot, official art, retro artstyle, sketch, lineart, and flat color do much of the work.

Mage's added tags lean every image modern. For an older look, use retro artstyle, 1990s \(style\), or anime screenshot, plus a sentence about the era's palette and shading, and expect a cleaner result than a true period piece.

Text is weak by design: a single word sometimes works, a short phrase now and then, and anything longer fails. Mage's negative list also removes speech bubbles, text bubbles, and artist signatures, so comic balloons will not appear. Add lettering in an editor afterwards.