Form: The Typology of Reformatted Bodies

A person contains an enormous amount of information: facial features, bodily proportions, gestures, expressions, age, gender, movement, and countless individual differences that resist simple classification. Yet when we encounter a highly recognizable figure, we do not necessarily need to process all of this information.
We may recognize what this is before we recognize what kind of person is presenting it.
Form begins with this perceptual phenomenon.
Cosplay makes this process particularly visible. When different people embody the same fictional character, they may differ radically in age, gender, ethnicity, bodily proportions, appearance, and physical condition. Yet viewers may still recognize the same character almost immediately.
Recognition, therefore, does not require the complete person.
It requires only a limited set of visual information sufficient to trigger recognition.
Form examines how the human body can be compressed into such a recognizable form.
The project examines two related forms of compression.
The first is aesthetic compression.
Popular culture, anime, games, social media, beauty practices, and digital imaging continuously produce bodily forms that can be recognized, imitated, and reproduced. Enlarged eyes, narrower or sharper facial contours, smooth skin, and idealized bodily proportions are among the recurring visual characteristics within these systems.
The second is semantic compression.
A complex individual can be rapidly reduced to a limited set of recognizable information: clothing, hairstyle, color, silhouette, props, gesture, and bodily proportions.
Once these elements are sufficient for recognition, what we recognize may no longer be the complete person, but a recognizable form.
Form is based on photographs collected at real cosplay events.
These source photographs were not produced under a unified photographic environment. Focal length, shooting distance, lighting, exposure, and other photographic conditions varied and were not fully controlled.
Form is therefore not a fully controlled photographic experiment, but an observational study.
After capture, the original backgrounds were removed and the subjects were placed within a shared green image field. This green field was not part of the original photographic environment. It was introduced afterward as a common image condition.
The operation removes the different environments in which the photographs were originally made, leaving the subjects and the visual information through which they can be recognized.
Form therefore asks:
Even when photographic conditions vary, can radically different bodies still be recognized as the same form through a limited set of visual information?

The subjects’ poses, gestures, and expressions remain.
They are not treated as photographic errors.
A cosplayer does not simply reproduce a character; they interpret that character through their own body. These differences therefore remain part of the work.
What is standardized is the image space, not the body.

AI: A Second Transformation and Mechanism of Aesthetic Compression
AI introduces another method of reformatting the body.
If cosplay translates the visual form of a fictional character into a human body, generative AI translates this already-characterized body back into an image.
This process is not simply reproduction.
In some generated results, visual characteristics strongly associated with anime aesthetics and contemporary digital portraiture appear to be further amplified: enlarged eyes, narrower or sharper facial contours, smoother skin, and more idealized bodily proportions.
AI may therefore amplify an existing process of aesthetic compression.
If cosplay represents a first transformation—the translation of fictional form into the human body—AI can be understood as a second transformation:
The body is compressed again toward visual forms learned by the machine.

This creates a significant tension.
AI must preserve enough information for the character to remain recognizable. At the same time, it may modify or remove individual bodily characteristics, bringing the body closer to an idealized form.
AI therefore participates in two forms of compression simultaneously:
semantic compression — retaining information sufficient for recognition;
aesthetic compression — reducing individual variation and reinforcing recognizable ideal forms.
AI thus becomes more than a generative tool. It becomes another method through which Form examines how the human body can be reformatted.
These observations should nevertheless remain provisional. They do not yet establish a definitive mechanism of AI generation. Further comparisons across subjects, characters, models, and generation conditions are necessary to determine whether these aesthetic tendencies are consistent.
AI and Photographic Traces
The AI experiments have also produced another unexpected observation.
In some cases, the original person was never photographed in front of a green screen. The original background was simply removed and manually replaced with a green field before the image was given to a generative AI system.
Yet some generated results contain green spill, edge contamination, and variations in the green background that were not present in the original image.
These traces were not produced by the original photographic event.
They were generated afterward within a photographic situation that had never physically occurred.
This cannot be directly interpreted as evidence that AI “understands” green-screen photography, nor can green spill be treated as proof of human presence.
It does, however, raise a question worth further investigation:
Does generative AI reconstruct not only a person, but also the visual conditions it has learned to associate with a plausible photograph?

The study of bodily form therefore begins to touch another question:
If an image can generate traces of a photographic event that never happened, what does photographic “evidence” mean?
From Diversity to Difference
Form is therefore not simply a typology of different cosplayers.
The inclusion of people of different ages, genders, ethnicities, and bodily conditions is not merely an attempt to display diversity.
What matters is:
Maximizing bodily difference while maintaining recognition of the same form.

The matrix therefore does not simply ask:
“How different are these people?”

It asks:
“How different can the bodies become before we stop recognizing them as the same form?”

This is also why the subjects’ poses and bodily performances should not be completely standardized.
If every person adopted exactly the same pose, the work would increasingly become a comparison of bodies under identical conditions rather than an investigation of recognition.
Form preserves a more difficult condition:
the image conditions can be controlled, while the body must remain a body.
Conclusion
Form is ultimately not a work about cosplay, anime, or AI.
These are the means through which the replacement of the individual by form becomes visible.
Photography has always selected, isolated, standardized, and translated the human body. Popular culture directs bodies toward recognizable and reproducible aesthetic forms; viewers extract the information necessary to recognize a character; and AI may further regenerate and amplify these existing aesthetic and semantic patterns.
The body therefore passes through successive transformations:
Fictional form → human body → perceptual recognition → machine generation → new form

Form examines this process.
It ultimately asks:
How much of a person can be transformed, standardized, or removed before we stop recognizing the person and begin recognizing only the form?

《Form：重新格式化身體的類型學》
一個人包含大量資訊：臉部特徵、身體比例、姿勢、表情、年齡、性別、動作，以及無數難以被分類的個體差異。然而，當我們面對一個具有高度辨識性的形象時，我們並不一定需要處理這些完整資訊。
我們可能先辨認出「這是什麼」，而不是「這是一個什麼樣的人」。
Form 從這個觀看現象開始。
Cosplay 使這個過程變得特別清楚。當不同的人扮演同一個虛構角色時，他們可能在年齡、性別、種族、身體比例、外貌與身體條件上存在巨大的差異；然而，即使這些差異非常明顯，觀看者仍可能在很短的時間內辨認出同一個角色。
這表示，角色的辨識並不需要完整的人。
它只需要一組足以觸發辨識的視覺資訊。
因此，Form 關注的是人體如何在觀看過程中被壓縮成一個可以被辨認的形式。
作品將這種壓縮分為兩個彼此相關的層次。
第一個是審美壓縮（aesthetic compression）。
流行文化、動漫、遊戲、社群媒體、美容實踐與數位影像，不斷產生可以被辨認、模仿與複製的身體形式。大眼睛、收窄或尖銳的臉部輪廓、平滑皮膚、纖細或理想化的身體比例，只是其中一些反覆出現的視覺特徵。
第二個是語義壓縮（semantic compression）。
一個複雜的人可以被觀看者迅速減縮為少數具有辨識作用的資訊，例如服裝、髮型、色彩、輪廓、道具、姿態與身體比例。
當這些資訊已經足以讓我們辨認角色時，我們所辨認的可能已經不再是完整的「人」，而是一個可以被辨識的形式。
Form：觀察
Form 取材自真實 cosplay 活動中的人物。
這些原始照片並不是在統一的攝影環境下完成的。不同人物的焦段、拍攝距離、光線、曝光與其他攝影條件並未完全受到控制。
因此，Form 並不是一個嚴格控制的攝影實驗，而是一項觀察性研究。
拍攝完成後，原始背景被移除，人物被重新置於共同的綠色色域中。這個綠色色域不是原始拍攝現場的一部分，也不是用來模擬某個具體世界，而是一個被後製建立的共同影像空間。
這個操作使原本不同的攝影環境退出畫面，留下人物以及使人物得以被辨認的視覺資訊。
因此，Form 所觀察的是：
即使攝影條件存在差異，具有巨大身體差異的人，是否仍然可以透過有限的視覺資訊，被辨認為同一個形式？

人物的姿勢、表情與動作則被保留下來。
它們不是攝影上的誤差。
Cosplayer 並不只是複製角色，而是透過自己的身體重新詮釋角色。因此，作品不試圖消除這些差異。
被統一的是影像空間，而不是人體。

AI：第二次轉化與審美壓縮
AI 為 Form 提供了另一種重新格式化人體的方法。
如果 cosplay 是將一個虛構角色的視覺形式轉譯進真人身體，那麼生成式 AI 則將這個已經被角色化的身體再次轉譯成影像。
這個過程並不只是複製。
在部分生成結果中，可以觀察到與動漫美型以及當代數位人像高度重疊的視覺特徵被進一步強化：眼睛變大、臉部輪廓更加收窄或尖銳、皮膚更加平滑，身體比例也可能向更加理想化的形式靠近。
因此，AI 可能使原本存在於文化中的審美壓縮進一步被放大。
如果 cosplay 是第一次轉化——將虛構形式壓入人體——那麼 AI 可以被理解為第二次轉化：
人體再次被壓縮到機器所學習到的視覺形式之中。

這產生了一個重要的矛盾。
AI 必須保留足夠的資訊，使人物仍然能夠被辨認為同一個角色；但同時，它也可能修改或刪除原本屬於個人的身體特徵，使人體更加接近某種理想化形式。
因此，AI 同時參與了兩種壓縮：
語義壓縮——保留足以辨認角色的資訊；
審美壓縮——減少個體差異，強化可辨識的理想形式。
AI 因此不只是 Form 的生成工具，也成為觀察人體如何再次被格式化的一種方法。
但這些變化目前仍應被視為觀察結果，而非對 AI 內部機制的最終結論。不同人物、角色、模型與生成條件是否產生一致的審美收斂，仍需要進一步比較與驗證。
AI 與攝影痕跡
AI 的實驗也產生了另一個意外的觀察。
部分原始真人照片從未在綠幕前拍攝。人物只是先被去除原背景，再人工置換成綠色色域，之後才交由生成式 AI 重新生成。
然而，在部分 AI 結果中，卻出現了綠色溢色、邊緣污染，以及與原始影像不同的綠色背景變化。
這些痕跡並不是原始攝影事件留下的。
它們是在一個實際上從未發生的攝影情境中，被後來生成出來的。
這不能被直接解釋為 AI 「理解」了綠幕攝影，也不能把綠色溢色視為真人存在的證據。
但它提出了一個值得持續研究的問題：
生成式 AI 是否不只重新生成一個人物，也同時重新建構它所學習到的「合理照片」應該具有什麼樣的視覺條件？

因此，Form 對人體形式的研究也開始觸及另一個問題：
當一個影像可以產生一個從未發生過的攝影痕跡時，我們又應該如何理解攝影中的「證據」？
從多樣性到差異
Form 因此並不是單純收集不同 Cosplayer 的類型學。
選擇不同年齡、性別、種族與身體條件的人，也不只是為了展示「多樣性」。
真正重要的是：
最大化人體之間的差異，同時維持同一形式的辨識。

因此，矩陣真正提出的不是：
「這些人有多不同？」

而是：
「人體可以有多不同，而我們仍然把它們辨認為同一個形式？」

這也是為什麼人物的姿勢與身體表現不應該被完全標準化。
如果所有人都採用完全相同的姿勢，作品所測量的就不再只是「形式辨識」，而會逐漸變成「在相同姿勢下的人體比較」。
Form 要保留一個更困難的條件：
影像條件可以被控制，但人體仍然必須作為人體存在。
結語
Form 最終並不是一個關於 cosplay、動漫或 AI 的作品。
這些只是使「形式如何取代個體」變得可見的媒介。
攝影一直在選取、隔離、標準化與轉譯人體。流行文化將人體導向可以被辨認與複製的美學形式；觀看者再從複雜的人體中抽取足以辨認角色的資訊；而 AI 則可能進一步把這些已經存在的審美與辨識模式重新生成，甚至加以強化。
人體因此經歷了一次又一次的轉化：
虛構形式 → 人體 → 觀看辨識 → 機器生成 → 新的形式

Form 所研究的，正是這個過程。
最終作品提出一個問題：
一個人可以被轉化、標準化、刪除多少，直到我們不再辨認「這個人」，而只辨認「這個形式」？