Li Yi-Fan’s site-specific installation Screen Melancholy brings together fragments of customised sculptural limbs with an uncanny yet humorous monologue film, featuring a naked human body as a puppet-character. Responding to the history and architecture of Venice’s Palazzo delle Prigioni, a former prison, the installation places audiences within a theatrical stage set. In a world of constant productivity, information overload and the necropolitics of war, Screen Melancholy confronts viewers with a sense of uncertainty and non-direction. Through fragmented bodies, accelerated speech and a strange mixture of humour and vulnerability, the work asks how we navigate a world increasingly mediated by screens, platforms and image-making technologies.

Jessica Wan: Screen Melancholy is a title that feels emotionally loaded and also points to the question of whose melancholy it is. What does melancholia mean to you and why did it become the emotional framework for this project? Li Yi-Fan: The title actually came from discussion with Raphael Fonseca, the curator. Before I made this work, we were looking at the work I had made previously and he found this kind of melancholic tone running through much of my practice.

I think he was referring to something that appears repeatedly in my work: this desire to know the world, to provide so much knowledge, while at the same time struggling with how much knowledge a person can actually deal with. The character in my films always wants to know more and deliver more, but there is also a sadness in that. Human beings are quite vulnerable when confronted with all this knowledge. Life is too short and the world is too big.

There is always this struggle between the individual and the world, especially in the age of the screen. We use screens so much. At first, the screen seems to open a window onto the world. But as we grow up, we start to realise that the screen is actually a flat surface with its own mechanisms behind it. It is not a transparent window through which we can simply understand the world.

For me, Screen Melancholy is about that relationship between the individual and the screen: how we approach the world through the screen but also how we begin to recognise the mechanisms behind it.

JW: Your work constantly unsettles distinctions between the physical and the digital, the original and the copy, the puppet and the puppeteer. Today, we’re surrounded by endlessly reproducible images, AI-generated content and simulated environments. Do those distinctions still matter to you or has authenticity become a less useful question? LYF: I think this is a really interesting – and also terrifying – moment to live through. One of the things people ask on platforms and social media has almost become a meme: “Is it real or is it AI?” The boundary between virtual and physical, between real and false, has become very blurry.

For me, though, it is not even the question of whether something is physical or digital, real or fake, AI or not AI. The more important question is how we recognise how an image was delivered – what the mechanics behind the image are.

Image-making has never been neutral. If you trace its history, you move from cameras to film, then computer graphics and now AI generation. There has always been a tension around how an image was created and what technology sits behind its production.

I think it is becoming increasingly important to recognise that technology, because we tend to take images for granted. It is very difficult to know what is actually happening behind the creation of an image. What matters more is understanding the technology behind the production.

I also think about this through my own relationship with AI. At first, I didn’t enjoy AI image generation at all. The initial text-to-image model didn’t really fit my practice. I come from a painting background, so I don’t naturally think through text. The process felt like finding the correct answer: finding the right latent space for the text and matching it with an image. But for me, image-making has never really been about finding the correct answer.

The technology has become much more sophisticated, though. It is no longer simply text-to-image; you can control an image through another image and the pipelines have become much more creative. If you look at the history of image-making technology, people have always taken new technologies and incorporated them into older pipelines to create new tools.

At the same time, there are serious questions about the data these systems are built from. AI models are trained on enormous datasets, much of which was collected without consent. Many of us may be contributing data without even knowing it. So the relationship between an AI-generated outcome and its source data becomes a very difficult question.

Even now, AI methods are embedded into software that we use every day. Something that used to require manual work can now be automated. It is convenient but there is always a moral question attached to the method. We need to keep discussing that.

JW: Both you and Raphael Fonseca have spoken about belonging to a generation that witnessed the shift from analogue media to the internet. Unlike younger generations, you remember images before they became endlessly networked. How has living through that transition – from VHS and DVDs to YouTube, social media, 3D printing and AI – changed the way you think about moving images and storytelling? LYF: I always think my generation was a particularly exciting generation, although maybe every generation feels that way. Computers didn’t really exist in my life until I was about seven and then suddenly they became a very important part of my life.

Growing up in Taiwan, the internet became a window onto the world. But we still have memories from before computers. VHS and DVDs played a big role in my life. I watched a lot of VHS when I was young and I have very precious memories of going into a VHS shop and looking for something I wanted without necessarily knowing what I was looking for.

That experience is very different from social media now. Social media always provides the things you want. If something appears in your feed, it means the system thinks you need it. We lose that experience of navigating – and especially navigating with a cost. When you rented a VHS, you paid for it. Every choice had a consequence, so you had to be conscious of what you picked. Now you subscribe and get everything, including things you don’t want.

I also experienced the beginning of YouTube. Around 2007, the algorithm was very different. You would get very random things and sometimes you had absolutely no idea why something appeared. I describe myself as part of the first generation of online learning. I learned most of my skills from YouTube. In the beginning, you could easily find one-hour tutorials. Now tutorials have become five minutes, one minute. You can really feel the pace of online platforms speeding up.

Speeding up has definitely changed how I think about storytelling. In my own films, the characters often speak very quickly and try to provide huge amounts of information or knowledge. It gives the audience a sense of anxiety. I feel a lot of anxiety within this sped-up platform environment. I enjoy it and hate it at the same time.

JW: You originally trained as a painter yet your practice now spans animation, game engines, performance capture and generative imagery. Was there a moment when you realised these digital tools could become your artistic language? LYF: I studied painting during my bachelor’s degree but during my training I found that I didn’t really enjoy the old-school, realistic approach. I spent a lot of time in the DVD library, and that was when I became much more interested in film and moving image. At one point, I really wanted to become a director.

I soon understood that becoming a director meant working with a large production team. I’m not very good at dealing with big productions involving so many people or with the hierarchy of the film industry. So animation seemed like a way for me to produce something independent.

I started making animation, including clay controlled by servo motors [physical motors that allow for precise control of position and speed]. That was about 10 years ago. It’s funny because, after 10 years, I’ve moved from all those analogue tools into digital ones but the principle is actually quite similar. It’s about building an environment, building the mechanics and then using them for performance.

I would say I became fully digital around 2018. Around that time, digital tools started to feel like they offered an individual the possibility of expressing themselves. Blender became very popular as open-source 3D software, and there were so many resources and tutorials on YouTube showing people how to build their own animations.

At the same time, [computer graphics game engine] Unreal Engine started to have the capability to create films and moving images. So around 2018, I started building my own toolkit inside Unreal. I’m not sure if you’re familiar with the term “machinima” – using a game engine to make a film. I would say my practice is largely based on this idea of machinima, between performance and animation.

What I enjoy about technology is the process. Animation and filmmaking are often considered big productions involving hundreds of people. I enjoy making the details myself. I need to understand how something works by actually making it. It is not purely about production; it is also about learning and craftsmanship.

JW: The film’s puppet character is unsettling because it’s simultaneously vulnerable, comic and strangely human. In this work, the body appears fragmented, damaged and manipulated, yet it also becomes a narrator and guide. How do these characters emerge? Do they begin with writing, drawing, performance or something else entirely? LYF: The character developed very slowly. I first used it in 2020, when it only had a head. Year by year, it gradually developed a body.

In the beginning, it was a 3D scan of my own body, so it has always been an avatar of me. It speaks in a strange tone, especially in Mandarin. I didn’t know how to animate the voice, so I kept my tone very flat and tried to match it to the character’s movements. Eventually, it became this strange combination of being stupid and smart at the same time.

He is always aware of the audience, almost like a lecturer or guide talking directly to you. One of my main references is the YouTube tutorial. I learned so much from these videos, where the narrator tells you how to make something, what is right and what is wrong. I enjoy that relationship of being guided but also questioning whether you should trust the person speaking.

The character has that same quality. He can be funny and confident but also unreliable and vulnerable. I think that tension makes the viewing experience intriguing.

JW: One of the most compelling aspects of your films is that they often feel like instructional videos that gradually unravel into existential monologues. You’re explaining animation software one moment and reflecting on consciousness the next. What interests you about borrowing the language of tutorials, gaming culture and internet pedagogy? LYF: Software tutorials are a major inspiration for me but they are also connected to the way I work. Because I work with machinima, voice acting becomes a performance. I sit in front of the computer and speak to the camera.

I also write a lot of notes during production. Whenever I learn something, I write it down. When I begin a new work, I go back through those notes and make connections between what matters to me and what I want to share. Then I turn them into a very minimal script and perform it.

The process is quite intuitive. I perform again and again until it feels right. Because everything is made in a game engine, I don’t have to wait for rendering, so I can immediately see the result and try again. This allows me to build the narrative step by step.

I can even use my phone to capture my expression and audio, and see the final result in real time. It almost feels like looking into a mirror while performing these strange movements and doing the voice acting.

JW: Digital technologies promise frictionless efficiency, yet they’re also sustained by vast infrastructures, extractive economies and growing environmental costs. Is that ecological dimension something you think about while making the work? LYF: I think this question has become increasingly difficult for me. When I first started using game engines, I wanted to build a DIY toolkit as an alternative to the film industry. I knew every detail of the puppetry tool I built and it made the practice feel very personal.

AI has changed my position. I spent years building my own tools and now almost anyone can generate a video in seconds through a platform. At the same time, our relationship with software has changed. In the past, you could buy software and use it forever – you could even give it to your son or grandson. Now everything is subscription-based. For digital artists, it can feel as though your ability to express yourself is being rented from someone else.

AI adds another layer because even with open-source models, the data behind them can remain a mystery. So I’m increasingly interested in the economic and technological systems behind image-making, not only the images they produce.

But I’m quite optimistic about image-making because human beings get bored very easily. When I first saw AI-generated images, I was shocked and excited but I got bored very quickly. Now I can already feel a kind of boredom inside many of these images. At the same time, it’s terrifying to think about all the data centres being built and what that means for the future.

JW: Your work often embraces uncertainty rather than resolution. After Screen Melancholy, where do you feel your curiosity is leading you next? Are there new questions – or perhaps new technologies – that you’re hoping to wrestle with? LYF: I think one of the things leading my curiosity now is the ethical problem of AI tools. Most of my work has focused on the instrumentality of image-making, and AI is a tool that I still haven’t fully navigated. I would like to spend more time working with these tools, to find a space between different tools and platforms.

Recently, I’ve become very interested in reading the licence agreements and privacy policies of different platforms. Whenever you generate something, you have to agree to these terms, and I find it interesting to see how they position artists and imagine what we might want to do with these tools.

Sometimes you find quite funny limitations. For me, these rules reveal how the platform is structured and the position it expects you to take. I don’t know exactly where the work will go next but I’m curious about these spaces between the artist and the systems that shape image-making.

---

李亦凡的場域特定裝置作品《鬱卒的平面》，結合了經過改造的雕塑肢體碎片，以及一段詭異卻幽默的獨白電影，片中一具赤裸人體化身成為木偶角色。作品回應威尼斯舊監獄普里奇歐尼宮邸的歷史與建築，將觀眾置於一個宛如劇場舞台的場景之中。在一個不斷追求生產力、資訊超載，以及戰爭的死亡政治無處不在的世界中，《鬱卒的平面》讓觀眾直視不確定性與無方向感。作品透過破碎的身體、急促的語速，以及幽默感與脆弱感交織的奇異氛圍，探問在日益由螢幕、平台及影像製作技術所調解的世界裡如何找到自己的方向。

Jessica Wan：《鬱卒的平面》這個標題帶有濃厚的情感意味，同時也讓人思考到底是誰在鬱卒。對你而言，鬱卒是甚麼？為甚麼它會成為這個項目的情感框架？李亦凡：這個標題其實是在與策展人哈法艾爾．馮希卡討論時想出來的。在創作這件作品前，我們回顧了我過往的作品，他發現這種鬱卒的調子貫穿了我不少創作。

我覺得他指的是我作品中反覆出現的傾向：渴望認識這個世界，提供大量知識，卻同時苦於思考一個人究竟能夠承受多少知識。我電影中的角色總是想知道和傳遞更多知識，但這種渴望同時帶有一份悲傷。面對龐大的知識，人其實相當脆弱。生命太短，而世界太大。

個人與世界之間總是存在這種掙扎，尤其因為我們正處於螢幕時代，頻繁使用螢幕。起初，螢幕似乎為我們打開了一扇通往世界的窗。但隨著年歲漸長，我們開始意識到螢幕其實只是一個平面，背後有其自身的運作機制。它並不是一扇透明的窗，讓我們可以輕易理解世界。

對我而言，《鬱卒的平面》探討的正是個人與螢幕之間的關係：我們如何透過螢幕認識世界，又如何逐漸察覺螢幕背後的運作機制。

JW：你的作品不斷動搖實體與數碼、原作與複製品、木偶與操偶者之間的分別。我們如今身處一個充斥著可無限複製影像、人工智能生成內容及模擬環境的世界。這些分別對你來說仍然重要嗎？抑或探討真實性已沒那麼有意義？李亦凡：我覺得這個時代非常有趣，同時也令人害怕。人經常在平台和社交媒體上問：「這是真的還是AI生成的？」，甚至幾乎變成迷因。虛擬與實體、真實與虛假之間的界線，已經變得非常模糊。

不過，對我而言，問題不在於東西究竟是實體還是數碼、真實還是假造、是AI還是不是，更重要的是我們如何辨識影像是透過甚麼方式呈現，以及影像背後有甚麼運作機制。

影像製作從來都不是中立的。追溯歷史就知道，影像製作技術從相機發展至電影，到電腦圖像，再到今天的AI生成技術。影像如何製作，以及背後採用了甚麼技術，一直存在某種張力。

我覺得辨識這些技術的愈來愈重要，因為我們往往把影像視為理所當然。要了解影像生成過程中實際發生的事非常困難，所以比起影像本身，更重要的是理解製作背後的技術。

我也會從自己與人工智能的關係去思考這個問題。起初，我一點也不喜歡人工智能影像生成，早期的文字生成影像模型並不適合我的創作。我接受的是繪畫訓練，所以我不會自然透過文字思考。整個過程感覺就像是在尋找正確答案：為文字找出合適的潛空間，再讓它與影像配對。但對我而言，影像創作從來都不是為了尋找正確答案而存在。

不過，這個技術已經成熟了很多。它不再局限於文字生成影像，還可以利用另一幅影像控制影像的生成，而整個製作流程也變得更有創意。如果你回顧影像製作技術的歷史，你會發現人們一直都將新技術融入舊的流程，從而創造出新的工具。

與此同時，這些系統所依賴的數據也引發了嚴肅的問題。人工智能模型以龐大的數據集進行訓練，其中許多數據是在未經同意的情況下收集，很多人甚至可能在毫不知情下提供了數據。所以人工智能生成的成果與源頭數據之間的關係，是一個相當棘手的問題。

即使時至今日，我們每日使用的軟件也已經內置了人工智能技術，過去需要人手處理的工作如今都可以自動化處理。這固然方便，但採用這種方法始終涉及道德問題，所以我們必須持續討論這些議題。

JW：你與哈法艾爾．馮希卡都曾提到，你們屬於親歷從舊媒體轉向互聯網的一代。與年輕一代不同的是，你記得的影像是在它們未在網絡上無限流通之前。親身經歷從VHS錄影帶和DVD影碟，到YouTube、社交媒體、3D打印及人工智能這些轉變，如何改變你對動態影像與敘事的看法？李亦凡：我一直覺得我這一代是特別令人興奮的一代，雖然或許每一代人都會這樣覺得。電腦大約在我七歲時才真正進入我的生活，然後突然成為生活中非常重要的一部分。

我在台灣長大，互聯網成為了我認識世界的一扇窗，但我們仍然保有電腦出現之前的記憶。VHS錄影帶和DVD在我的成長過程中扮演了很重要的角色，小時候看過很多VHS錄影帶，時至今日，我仍然非常珍惜那些走進錄影帶店尋找想看的東西但其實不知道自己究竟想找甚麼的回憶。

這種體驗與今日的社交媒體截然不同。社交媒體總是會給你看你想要的東西，如果某些內容出現在你的動態消息中，就代表系統認為你需要它，我們因此失去了自行探索的體驗，尤其是那種涉及成本的探索。租借VHS錄影帶要付錢，每個選擇都有後果，所以你必須清楚自己選擇了甚麼。現在只要訂閱服務就可以獲得所有內容，包括自己根本不想看的東西。

我也經歷了YouTube剛起步的時期。大約在2007年的時候，演算法與現在非常不同。你會看到各式各樣、毫無規律的內容，有時甚至完全不知道為甚麼某段影片會出現在眼前。我會形容自己屬於第一代透過網絡自主學習的人，大部分技能都是從YouTube學會。早期很容易就能找到長達一小時的教學影片，現在的教學影片變成五分鐘，甚至一分鐘，你可以確實感覺到網絡平台的節奏正在加快。

這種加快的速度無疑改變了我對敘事的想法。在我自己的電影中，角色語速往往飛快，嘗試傳遞大量資訊或知識，令觀眾感到焦慮。在這種節奏加快的平台環境中，我自己也感到強烈焦慮，既享受又厭惡。

JW：你原本接受繪畫訓練，但如今的創作實踐涵蓋動畫、遊戲引擎、動作捕捉及影像生成。你有沒有在哪個時刻意識到，這些數碼工具可以成為你自己的藝術語言？李亦凡：我在學士學位期間主修繪畫，但在受訓的過程中，我發現自己並不喜歡傳統的寫實繪畫方式。我花很多時間在DVD資料室，也正是在那段時間，我開始對電影和動態影像產生更大的興趣。有一段時間，我真的很想成為導演。

但我很快就明白，成為導演意味著要與一支龐大的製作團隊合作。我不太擅長處理多人的大型製作，也不喜歡電影產業的階級制度，所以動畫似乎是一種能夠讓我獨立創作的方式。

大約十年前，我開始製作動畫，包括利用〔能精準控制位置和速度的實體摩打〕伺服摩打控制黏土。有趣的是，十年過去，我從那些舊工具轉向數碼工具，但兩者的基本原理其實很相似：先建立環境，構建運作機制，再利用這些機制實現表演。

我會說自己大約在2018年全面轉向數碼創作，在那段時間，數碼工具開始讓我感覺到個人也能透過工具表達自己。開源3D軟件Blender變得非常流行，而YouTube上也有大量資源和教學，教人如何製作自己的動畫。

與此同時，〔電腦圖像遊戲引擎〕虛幻引擎也開始具備製作電影和動態影像的能力，所以大約在2018年，我開始在虛幻引擎裡建立自己的工具組。不知道你有沒有聽過「引擎電影」這個詞，它的意思是利用遊戲引擎製作電影。我會說自己的創作很大程度上是建基於引擎電影的概念，介乎表演與動畫之間。

我喜歡科技的地方，在於它的製作過程。動畫和電影製作往往需要數百人參與大型製作，但我喜歡親手處理細節。我要親自動手製作才能真正理解東西的運作。這不只是製作成品的過程，也關乎學習與工藝。

JW：電影中的木偶角色令人不安，因為它同時脆弱、滑稽，又帶有一種奇異的人性。在這件作品中，身體呈現出破碎、受損和被操控的狀態，卻同時是敘述者和導覽員。這些角色是如何形成的？它們最初源自文字、繪畫、表演，還是其他不同的東西？李亦凡：這個角色是經過很長時間逐漸發展出來的。我在2020年首次使用它，當時它只有一個頭。之後每一年，它的身體都會慢慢長出。

最初它是以我的身體進行3D掃描而成，所以它一直都是我的化身。它說話的聲調很奇怪，尤其是說普通話時。我當時不知道如何為聲音製作動畫，所以我維持非常平板的語調，再嘗試配合角色的動作。最後，它逐漸變成一種奇怪的結合，既愚蠢又聰明。

它一直知道觀眾的存在，就像是講師或導覽員直接向你說話一樣。我其中一個主要的參考來源就是YouTube教學影片。我從這些影片學到很多，旁白會告訴你如何製作某樣東西、甚麼是對的、甚麼是錯的。我很喜歡這種被引導的關係，但同時也會質疑自己是否應該相信說話的人。

這個角色也有同樣的特質，它可以很有趣、很有自信，卻同時不可靠和脆弱。我認為正是這種張力，令觀看體驗變得引人入勝。

JW：你的影片其中一個最引人入勝的地方，在於它們往往像教學影片，卻逐漸演變成探討存在問題的獨白。前一刻你還在解釋動畫軟件，下一刻已經轉而思考意識。借用教學影片、遊戲文化及網絡教學的語言，對你而言有甚麼吸引之處？李亦凡：軟件教學影片是我的一大靈感來源，但它們也與我的創作方式有關。我用引擎電影的方式創作，所以配音本身就會成為一種表演，我會坐在電腦前對著鏡頭說話。

在製作過程中，我也會寫下大量筆記。每當學到新東西，我都會記錄下來。開始一件新作品時，我會重新翻閱這些筆記，將自己在意的事情與想分享的內容連結起來，再把它們整理成一份精簡的劇本，然後親自演繹。

這個過程很依靠直覺，我會反覆表演，直到感覺良好為止。由於所有內容都是在遊戲引擎中製作，我不用等渲染完成，就能立即看到結果並再試，讓我可以一步一步建構敘事。

我甚至可以用手機捕捉自己的表情和聲音，並即時看到最終效果。整個過程幾乎就像一邊看著鏡子，一邊做出奇怪的動作和進行配音表演。

JW：數碼技術保證帶來毫無阻力的高效率，但其運作同時依賴龐大的基礎設施、資源掠奪式經濟，以及日益沉重的環境成本。你在創作時會思考這些生態層面的問題嗎？李亦凡：我覺得這個問題愈來愈難回答。最初開始使用遊戲引擎時，我想建立一套DIY工具組，作為電影產業以外的選擇。我很清楚自己製作的木偶工具的每個細節，令整個創作實踐變得非常個人。

人工智能改變了我的定位。我花了多年時間建立自己的工具，現在幾乎任何人都可以透過平台，在幾秒內生成一段影片。與此同時，我們與軟件的關係也改變了。以前你可以買下軟件永久使用，甚至把它傳給自己的兒孫，現在甚麼都變成訂閱制。對數碼藝術家而言，感覺像是自己表達創意的能力是向別人租借回來的。

人工智能又帶來另一重問題，即使使用開源模型，背後的數據來源仍可能是一個謎。所以我愈來愈關注影像製作背後的經濟與技術體系，而不只是這些體系所產生的影像。

不過我對影像創作仍然抱持相當樂觀的態度，因為人類很容易感到悶。第一次看到人工智能生成影像時，我既震驚又興奮，但很快便覺得沒趣，現在我已經對很多這類影像感到乏味。與此同時，一想到大量數據中心正在興建，以及這對未來的意味，又令人感到恐懼。

JW：你的作品往往擁抱不確定性，而不是追求明確的結論。完成《鬱卒的平面》之後，你覺得自己的好奇心會把你帶往哪個方向？有沒有哪些你希望接下來繼續探索和應對的新問題或科技？李亦凡：現在引領我繼續探索的其中一個方向，是人工智能工具所涉及的倫理問題。我過往大部分作品都聚焦於影像製作的工具性，而人工智能仍然是一種我尚未完全掌握的工具。我希望花更多時間使用這些工具，在不同工具和平台之間尋找空間。

最近，我對閱讀不同平台的授權協議和私隱政策產生了很大興趣。每當你生成某些內容，就必須同意這些條款。我覺得很有意思的是，這些條款如何定位藝術家，以及平台如何想像我們可能會怎樣使用這些工具。

有時，你會發現一些很有趣的限制。對我而言，這些規則揭示了平台的架構，以及平台預期你採取的立場。我不知道我的創作下一步會走往甚麼方向，但我很好奇藝術家與那些形塑影像製作的系統之間所存在的空間。