I am running a series of tests using AI coding applied to various topics in visual effects, animation and astronomy visualization. These little demonstrations are intended to be nothing more than proofs of concept.
Posts like this one will be collected and added to a blog devoted specifically to this topic, and related topics of using technology in visual effects, animation and some astronomy visualization. In the interim these posts will be added to my general blog.
Here is a list of the tests in progress:
1. Write a program that examines all movies in a folder, determines their resolution, and writes a script to transcode the movies to the highest resolution maintaining original quality, and assemble those clips into a single movie.
2. Write a program that runs on a smartphone to listen to a conversation, determine who is speaking, and notifies the speaker if they are overly dominating a conversation.
3. Write a program that uses the famous Yale Bright Star Catalog (YBSC) and visualizes it in a web browser using Web GL.
4. Write a browser based DAG (directed acyclic graph) designer that might be a front end to any application that could use such a graph. Traditionally this has been such things as a compositor, a material designer or an animation script.
5. Write a program that analyzes audio and visualizes the themes and instruments in a way that we might a click track for animation from that audio.
6. Write a program that analyzes the famous "find a grave" database, downloads the graves from a designated region and visualizes the result in a web browser.
The results so far have been promising. In all cases, certain aspects of the test application have been achieved in a few days at most. In at least one case, the audio cases in particular (e.g. 2 and 5 above), the limitations of my knowledge are exposed. The process of analyzing continuous speech to determine who is speaking is non-trivial. Same for the process of deducing themes in music. Nevertheless, in both cases, I quickly got to the place where I had a good framework for testing ideas and a GUI which, while certainly not final, was useful for prototyping.
As with using an AI image generation program, it is all about a carefully constructed prompt.
The feeling of using this AI coding is actually very positive. It is a very productive process. Whether it will result in final applications is to be determined.
Here are some examples from works in progress.
Audio analysis
Game grid
Diarization app for android
No comments:
Post a Comment