Can't seem to generate full body shots

[Karl_Manning]

Although it gets the setting reasonbly right, it almost refuses to draw any person in anything other than a portrait shot.
eg

the Mad Alchemist with curly-toed slippers on feet, wild hair, grey beard, piercing blue eyes, wearing a maroon rune-covered hooded cloak creating concoctions in the lab

Generates only portraits. The only way I can get a full person is to drop the length of the description down to something like

the Mad Alchemist with curly-toed slippers on feet, wild hair, grey beard, piercing blue eyes

Have tried prompts like "from the front" "full body" "showing feet" etc but for some reason it just ignores them.

Not just on this prompt - basically everytime tried to do a human "art" " graphic" it does it like this.

[campagnoli]

I had some success on some of the variations using "full shot" in the prompt. I tried 'uncropped' or "head to toe", but "full shot" had the best results.
And of course if you mention the shoes in the prompt, like "shinny boots" and set proportion to vertical/portrait works even better.

[campagnoli]

And of course if you mention the shoes in the prompt, like "shinny boots" and set proportion to vertical/portrait works even better.

[amies20281713]

It is nearly impossible for me to get it to show a standing person without distortion or a weird pose. I think different aspect ratios might help with this.

[droopydog500]

Hello @amies20281713,

Thank you for your message. I am sorry you are having this problem. 

Yes, portrait aspect ratio helps. I have found that when all else fails, drawing an outline of what you want and using that as a structure reference can do the job. Just do not use thin lines in the drawing, do not make the drawing transparent, and make sure your structure reference is the same aspect ratio as your generated image.

My best,
    droopy

[market23607553x8px]

人物描画で全身とプロンプトに書いている
和文での指示”全身を構図に入れる”"全身を描画"
英文での指示”Full-body composition””Frame-filling composition”
他に”爪先まで描画”の指示も書いた
カンバスは縦

足が見切れるものしか生成しない。
クレジットを消費するばかりで完全なイラストが出来ない。
Fireflyの翻訳はMicrosoftのものだそうで、それに問題があるかもしれないが、それ含めて製品・商品だ。
イラストの良し悪し以前に構図・画角が狂うのは致命的な欠陥。
無駄にクレジットを消費させるのは詐欺・搾取でしかない。

全身構図を求めていながら足が切れるのは絵として無価値だ、クレジット＝金銭であるのでAdobe社には早急に改善願う。

[droopydog500]

Hello @market23607553x8px,

Thank you for your message. I am sorry you are having this problem. 

This is actually a problem in English, too. There are lots of little things you can try, and you tried many already. I have found that the most effective Suggestion is to hand draw an outline of a person and use that as a structure reference. You do not have to be an expert artist or make the perfect drawing. You just need it to instruct the model on the perspective of the object you want in the generated image.

I hope this helps.

Thanks,
    droopy

[market23607553x8px]

教えてくれてありがとう

教えてくれた方法を試してみる。
写真を生成した場合足の形状が人間のものと見えないことが多々あるので意図して見切っているのかと疑いたくなる。
クレジットを使う製品・サービスなのでAdobeには早急に改善してもらいたいものです。

[droopydog500]

Hello @market23607553x8px,

If just the feet are bad, you could take that into generative fill to correct the feet.

My best,

    droopy

[Brett3798706766sz]

Same problem. It's almost as though they are preventing the AI from making full-body shots that include a face.

[market23607553x8px]

教えてくれてありがとう
試してみるよ。
他の生成AIで画像を修正させるのはFireflyが不完全ということで、Adobe社は重く受け止めて早急に改善してほしい。

[droopydog500]

Hello @market23607553x8px,

I agree. Trying to get a full body image is too hard. You should not have to do all these work arounds.

    droopy

[market23607553x8px]

？カンバスを変えることはとっくにやっています。
縦は3：4しかないが、それでも切れるのは単にAIの理解力が足りない＝商品価値が低いからです。
本来カンバスが横向き16：9でも背景が大きくなり人物すべてが描画されているのが正しく対価に値する。