[2025.09.04] 사전조사 및 교수 면담

상황 정리

I'm trying to create a software program that 

1) gets a prompt, interpret what it means

2) and create a short video that makes a character do an action related to the prompt

---

You're basically combining natural language understanding with generative media(video/animation). 

 

1. Natural Language Processing (NLP) / LLMs

- Learn how to parse promts into structured meaning

    ("make the character wave hello" -. {action: "wave", target:"character"})

- Topics:

    * Prompt enginerring

    * Fine-tuning, Instruction-tuning LLMs

    * Text-to-structured-data conversion

 

2. Computer Vision & Generative Models (Images/Video)

- Learn how generative AI creates images and video frames

- Topics:
    * Diffusion models (Stable Diffusion, Runway Gen-2, Pika Labs)

    * Video diffusion and frame interpolation

    * Motion transfer (making a static character perform a motion)

 

- Character Animation & Motion Models

    * Pose Estimation (OpenPose, MediaPipe) -> define character skeletons

    * Motion Capture Datasets -> Mixamo, HumanML3D

    * Text-to-motion modles -> MotionGPT, T2M-GPT, AnimateDiff, DreamPose

https://motion-gpt.github.io/

MotionGPT

https://www.youtube.com/watch?v=FNY6A0tFrms

 

---

자진프는 매주 레포트를 내야함
세 사람이 다르게 내지 않았으면 좋겠고, 그냥 세 사람의 파트가 보이도록, 하나의 통일된 레포트를 내라
3학점을 주는 기준이 레포트임
실제로 일을 했다는 증빙으로 쓸거임

승인을 몰아서 하는 편이라서 승인이 좀 걸릴수도 있다

on-demand 미팅 신청