TTXC 台灣文化科技大會 - AI進行式

interactive UI/UX ui design Exhibition Exhibition Design expo Event

AI進行式

展出地點：高雄流行音樂中心珊瑚礁群主場館2F

呼應今年最熱門關鍵字生成式AI(generative AI)，與台灣人工智慧實驗室合作，透過多項作品展示與互動，讓民眾了解AI如何辨識、理解及生成。其中包含「AI虛擬主播」人工智慧技術製造的虛擬人物，不僅擁有擬真外貌，更擁有自己獨特的個性與動作；「AI 夢想攝影棚」虛擬場景製造，只要觀眾擁有對場景的想像，就能帶領觀眾進入幻想世界；最後「「AI 你的名字」字體模擬系統，讓操作者只要手寫一個字，AI即能根據此字，模仿出完整筆跡，並顯示由這字體生成的名字解析。

AI浪潮襲來，台灣人工智慧實驗室從三項作品向民眾宣示AI正在我們生活中迅速發展，並想像未來AI將如何加速我們工作和生活的模式。

AI Ongoing

Location: VENUE 2F

In response to the most popular keyword of the year, generative AI, we work with the Taiwan AI Labs to help the public understand how AI identifies, understands, and generates. Works include the AI Virtual Newscaster, a virtual character made with AI technology with a lifelike appearance and its personality and movements. The AI Dream Studio creates virtual scenes. As long as visitors can picture it, it can lead them into the fantasy world. Last, it’s the AI Your Name font simulation system. The operator only has to write a word by hand and the AI will be able to imitate their writing based on it and display the name analysis generated from this font.

As AI becomes the trend, the Taiwan AI Labs demonstrate the rapid development of AI to the public through three works and pictures of how AI will be able to speed up our work and life.

展場設計

入口處設計一道隧道，穿梭光牆象徵進入新的世界作為空間切割，走廊空間說明了體驗流程、AI技術介紹⋯⋯等。

Exhibition Design A tunnel is designed at the entrance, serving as a symbolic gateway into a new world, demarcating the space. The corridor space illustrates the experiential process, introduces AI technologies, and so on.

「AI 你的名字」

民眾在平板上寫上自己的名字，AI 將為你生成你專有的字體，並顯示由你的字體生成的“名字解析”。

技術說明：

台灣人工智慧實驗室搜集大量不同字體的文字，輸入為一張基底字體文字的圖片A，以及一張隨機取樣的風格字體的圖片B(如觀眾隨機的手寫字)，訓練模型根據B的風格去改變A的字體形狀。

模型的架構為對抗式學習的架構，分為「生成器」以及「辨別器」，「生成器」像臨摹家；「辨別器」像評審。臨摹家根據目標字體改寫基底文字，評審則辨認生成出的文字是否屬於風格字體，臨摹家的目標是騙過評審使生成文字更像目標字體，評審會根據現有的字體資料去區分生成字以及目標字體，訓練過程中兩個模型交互影響成長，最終達到字體生成的效果。

未來應用：可以想像在一些設計軟體中，使用者如果想要某段文字用自己的手寫風格來展現，只要提供自己手寫的範例，系統就可以生成整段文字而且看起來就像使用者親手寫的一樣。或者在卡片、邀請函、海報等設計中，能夠快速模仿和生成不同風格的文字，增添個性和獨特性。

"AI Your Name"

Visitors can write their names on a tablet, and AI will generate a personalized font for you, displaying a "Name Analysis" created from your unique font.

Technical Explanation: The Taiwan Artificial Intelligence Lab collects a vast array of text in different fonts. It takes an image A of a base font text and an image B of randomly sampled style font (such as handwritten text by the audience). The model is trained to modify the shape of the base font A according to the style of B.

The architecture of the model involves Generative Adversarial Networks (GANs) comprising a "Generator" and a "Discriminator." The "Generator" acts like an imitator, modifying the base text according to the target font, while the "Discriminator" acts as a judge, determining if the generated text matches the style font. The goal of the Generator is to deceive the Discriminator to make the generated text more similar to the target font. The Discriminator distinguishes between generated and target fonts based on existing font data. During the training process, these two models interact and influence each other's growth, ultimately achieving the desired font generation effect.

Future Applications: This technology can be envisioned in design software where users, wanting a segment of text to reflect their own handwriting style, can provide a handwriting sample. The system can generate entire paragraphs of text that appear as if the user wrote it themselves. Additionally, in designs such as cards, invitations, posters, etc., this technology can quickly imitate and generate text in various styles, adding personality and uniqueness.

「AI你的名字」成果顯示畫面

「AI虛擬主播」

對著麥克風與AI虛擬主播講話，AI主播就會跟你對話。台灣文化科技大會引領全球首次使用「AI未來攝影棚」生成場景、文字、音樂與數位代言人- AI虛擬主播「艾雅婷」！

技術說明：

「AI虛擬主播」主要結合兩項AI應用：AI Avatar虛擬人像及AI語音合成。

AI虛擬主播是透過人工智慧技術製造的虛擬人物，擁有自己獨特的個性、動作和擬真外貌，經過大型語言模型訓練後，能夠以多種語言流利地與觀眾互動，並具備高度自動化的表演能力。

AI語音合成使用 AI 技術將文字合成自然流暢的人類語音，台灣人工智慧實驗室先進的技術能在生成語音時保持自然流暢性和聲音的真實性，聽到最在地且最自然的台灣口音。

未來應用：可以用於電子書朗讀、語音導航、語音助手、戲劇和電影配音、客服應用、教育與培訓、訪問輔助、廣告行銷。

"AI Anchor"

Speak into the microphone, and the AI anchor will engage in conversation with you. The Taiwan Culture Tech Expo pioneers the world's first use of the "AI Future Studio" to generate scenes, text, music, and digital spokesperson - the AI anchor "i-Yating"!

Technical Explanation: The "AI anchor" integrates two AI applications: AI Avatar for virtual human images and AI Speech Synthesis.

The AI anchor is a virtual character created using artificial intelligence technology. It possesses its own unique personality, movements, and lifelike appearance. After extensive training with large language models, it can interact fluently with the audience in multiple languages and has highly automated performance capabilities. AI Speech Synthesis employs AI technology to synthesize text into natural and fluent human speech. The advanced technology from the Taiwan Artificial Intelligence Lab maintains natural fluency and authenticity in the generated speech, incorporating the most authentic and local Taiwanese accents.

Future Applications: It can be used for audiobook narration, voice navigation, virtual assistants, dubbing in theater and movies, customer service applications, education and training, interview assistance, and advertising and marketing.

「AI夢想攝影棚」

只要輸入你對場景的想像，如：「有滿滿的粉紅色棉花糖以及繽紛的氣球」，AI 便能在短短幾分鐘內，生成夢幻的沉浸式場景，帶您進入幻想世界。

體驗者可進入自己創造的夢幻攝影棚拍照、攝影、打卡。

技術說明：

生成式科技的 Stable Diffusion 技術已被廣泛應用於各種 AI 生圖領域。台灣人工智慧實驗室則基於Stable Diffusion 技術，賦予了AI創造空間的能力。我們運用 Dreambooth 技術來指導模型生成全景效果，將其用作虛擬攝影棚的一部分。

未來應用：

透過文字的描述，AI 能自動生成360環景角度的具有空間資訊且可編輯的場景，可應用在各種內容製作、降低佈景成本。

"AI Dream Studio"

Just input your imagination for a scene, like "filled with pink cotton candy and vibrant balloons," and AI can create a dreamy immersive setting within a few minutes, transporting you to a fantastical world. Participants can enter the dreamy studio they've envisioned to take photos, shoot videos, and capture memorable moments.

Technical Explanation: The Stable Diffusion technology in generative science has found widespread applications in various AI-generated image fields. Building upon Stable Diffusion, the Taiwan Artificial Intelligence Lab has empowered AI with the ability to create spaces. We utilize the Dreambooth technology to guide the model in generating panoramic effects, integrating it as part of the virtual photo studio.

Future Applications: By describing scenes through text, AI can automatically generate editable, 360-degree panoramic scenes with spatial information. This technology can be applied in various content productions, significantly reducing set design costs.