This is great - now I can get the authentic conference experience of a disengaged speaker reading out the slides in a monotone, without all the hassle of international travel and scheduling.

In all seriousness, there could be more utility in this if it helped explain the figures. I jumped ahead to one of the figures in the example video, and no real attention was given to it. In my experience, this is really where presentations live and die, in the clear presentation of datapoints, adding sufficient detail that you bring people along.

The samples from the authors' GitHub are just some text vomited onto slides, and the AI voice reading them point by point. Exactly the opposite of a good presentation.

Ah I guess if you’re very bad at presentations, then this could be beneficial. However, scientific presentations are meant to be communicating science and making things stick to your audience (no matter if it’s scientists or children you’re presenting to). This does not fix that problem at all. For anyone thinking of using this: please watch: https://m.youtube.com/watch?v=Unzc731iCUY and maybe a talk from Jane Goodall on how to engagingly show your science. I would hate to see a lot of conference presentations be made with this generator.

Another thing that improved my personal presentation skills was noting down why I liked a presentation or why I didn’t - what specific things a person did to make it engaging. Just paying attention to that improved my presentation skills enormously

Very interesting project, and I found two things particularly smart and well executed in the demo:

1. Using a "painter commenter" feedback loop to make sure the slides are correctly laid out with no overflowing or overlapping elements.

2. Having the audio/subtitles not read word-for-word the detailed contents that are added to the slides, but instead rewording that content to flow more naturally and be closer to how a human presenter would cover the slide.

A couple of things might possibly be improved in the prompts for the reasoning features, eg. in `answer_question_from_image.yaml`:

  1. Study the poster image along with the "questions" provided.
  2. For each question:
     • Decide if the poster clearly supports one of the four options (A, B, C, or D). If so, pick that answer.
     • Otherwise, if the poster does not have adequate information, use "NA" for the answer.
  3. Provide a brief reference indicating where in the poster you found the answer. If no reference is available (i.e., your answer is "NA"), use "NA" for the reference too.
  4. Format your output strictly as a JSON object with this pattern:
     {
       "Question 1": {
         "answer": "X",
         "reference": "some reference or 'NA'"
       },
       "Question 2": {
         "answer": "X",
         "reference": "some reference or 'NA'"
       },
       ...
     }

I'd assume you would likely get better results by asking for the reference first, and then the answer, otherwise you probably have quite a number of answers where the model just "knows" the answer and takes from its own training rather than from the image, which would bias the benchmark.

Hrhr, I'd love to have automatic CODE generation from Scientic Papers :D

Damn, they automated Károly Zsolnai-Fehér

Project page (links to both github and arxiv): https://showlab.github.io/Paper2Video/

This is the opposite of what I want. I'd rather turn videos into articles.