#ALM102dc Deepfake in Video Making

Posted byKent Patrick HuangPosted inUncategorized

Hey guys how’s it going? It’s been though situation for a while, and you might be stressed up a bit, but allow me to stretch your brain cells a bit more with something that you might want to take notes of which is deepfake.

Video Brief

For the final assignment, I’ve chosen AI in video making, specifically deepfake which is a booming topic in the recent years. The goal of the video is to give a neutral perspective of the technology to make people aware of its benefits and drawbacks and things to be cautious about. My intended audience is the general public since this technology is something that could affect anyone. The key takeaway is so that the viewer could be more aware with the spread of fake news. The call to action is for those who are really into deepfake detection system to read more about it and do their own research as it is really complex.

Written Reflection

My test began a month before the due date since deepfake is an uncharted territory for me. I practiced using DeepFaceLab which is the program I used for my deepfake test and tried several machine learning training methods before finding the best one which I applied in the first shot of the video, I had to do around 40 hours of training to get the desired result. I did my research a week prior to shooting, however since I had other assignments, I couldn’t focus entirely in it. The real research began 3 days before shooting. I planned to finish my research on Wednesday and then shoot at the same day which I successfully did thanks to Bob’s support and after days of extremely tiring research (since it’s quite hard to find the scholarly source for it). Shooting didn’t take long since I’m doing a normal analytical video, so I went straight to editing which took quite a while since it’s hard to find the required footages for my script. Hence why I used paint in some of my footage to compensate it since I don’t have an animation software. I used After Effects for motion graphics for some explanations to illustrate the point I’m trying to convey. I also used some memes in the video because it just pops in my head and it’s relevant to what I’m trying to say. Lo-fi music was used for the video to illustrate that this video is a non-formal video and also to make more intriguing to watch. Moreover, a new character appeared in my video which is none other than myself dressed in a jacket, an ushanka, sunglasses, and mask. To illustrate that it could be anyone behind the mask since the face is obscured.

Making a 7 minutes video about a new topic in the industry is definitely a challenge in itself. The lack of scholarly resource makes it hard for me to progress in the outset during my research process. But thanks to Bob’s help and support, I’m able to finish my video which I am keen on doing. Finding the source is one problem but finding the video as a b-roll is another one. During editing I noticed that some of the things I said have YouTube videos for it, however, most of them are copyrighted. This is one of the reasons why I prefer to shoot my own B-rolls for my projects, however, in this instance, it’s rather impossible to shoot my own videos for some parts of the video. The biggest problem was the part where I talked about Corridor Crew’s creation since all their works are copyrighted. And that’s why by using either paint or After Effects, I hope to compensate the gap I have by being as creative as I can be. Before I talked about the problems of deepfake, I changed from video footage to just texts. It’s not because of lazy editing, I aimed to draw the viewer attention as it is a contrast to the previous shots which portrays deepfake as something really good, my goal is to take the viewer’s attention so that they’re aware that deepfake is not always good, but it has its own dangers. The last challenge I had to face is my housemates which are getting rather noisy with their new switch game, I tried using NVIDIA’s artificial intelligence called RTX Voice which would isolate my voice better and remove all the noise that’s coming from any ambient sounds which produce a less distorted audio compared to adobe’s denoiser.

Reference List

  1. David 2020, Social Media Statistics Australia – January 2020, Social Media News, retrieved 27 May 2020, <https://www.socialmedianews.com.au/social-media-statistics-australia-january-2020/>
  2. Don’t Trust ANYTHING You See – Deepfakes explained 2019, Youtube, Techquickie, 1 November, retrieved 26 May 2020, <https://www.youtube.com/watch?v=4JNBDwd40is>
  3. DuPreez, H 2020, Introduction to Deepfake, developer.com, retrieved 17 May 2020, <https://www.developer.com/design/introduction-to-deepfake.html>
  4. Fake Videos of Real People 2018, Youtube, TED, 26 July, date retrieved 28 May 2020, < https://www.youtube.com/watch?v=o2DDU4g0PRo>
  5. Goggin, B 2019, From Porn to ‘Game of Thrones’: How Deepfakes and realistic-looking Fake Video Hit it Big, Business Insider, retrieved 26 May 2020, <https://www.businessinsider.com/deepfakes-explained-the-rise-of-fake-realistic-videos-online-2019-6?r=AU&IR=T>
  6. Guera, D & Delp, E J 2018, Deepfake Video Detection Using Recurrent Neural Networks, IEEE Xplore, retrieved 28 May 2020, <https://ieeexplore.ieee.org/abstract/document/8639163>
  7. New Technique for Detecting Deepfake Videos 2019, Youtube, UC Berkeley, 18 June, date retrieved 28 May 2020, < https://www.youtube.com/watch?v=51uHNgmnLWI>
  8. Paris, B & Donovan, J 2019, Deepfakes and Cheap Fakes: The Manipulation of Audio and Visual Evidence, Data & Society, retrieved 28 May 2020, < https://datasociety.net/wp-content/uploads/2019/09/DS_Deepfakes_Cheap_FakesFinal-1-1.pdf>
  9. Shao, G 2020, Fake Videos Could be the Next Big Problem in the 2020 Elections, CNBC, retrieved 27 May 2020, <https://www.cnbc.com/2019/10/15/deepfakes-could-be-problem-for-the-2020-election.html>
  10. Thomas, B 2020, Deepfakes: A Threat to Democracy or Just a bit of Fun, BBC, retrieved 27 May 2020, <https://www.bbc.com/news/business-51204954>
  11. Un0m 2019, With enough computer power…, Reddit, retrieved 28 May 2020,  <https://www.reddit.com/r/explainlikeimfive/comments/d7pcgm/eli5_what_is_the_danger_with_deepfakes/>

Creative Commons

Videos

  1. Amazing DeepFake Putin replaces Kevin Spacey Let me be Frank by Best of TV licensed under Creative Commons CC BY 3.0 [https://www.youtube.com/watch?v=8euP8iq6OCY]
  2. Deepfake: A Brief History of Unreliable Images by vlogbrothers licensed under Creative Commons License CC BY 3.0 [https://www.youtube.com/watch?v=dVa-00GYHas]
  3. Hands Typing on Laptop – Free HD Stock Footage (No Copyright) by Free Stock Footage 4K licensed under Creative Commons License CC BY 3.0 [https://www.youtube.com/watch?v=nXCI3F-zhxk]
  4. Keanu. by vlogbrothers Collider licensed under Creative Commons License CC BY 3.0 [https://www.youtube.com/watch?v=R0xDrYg-QJ4]
  5. RARE TUPAC SHAKUR IN THE STUDIO WITH BIG SYKE, RICHIE RICH. LIVE RECORDING. by NAPPYGOD X licensed under Creative Commons License CC BY 3.0 [https://www.youtube.com/watch?v=OqzNY0CCRIk]
  6. Reading a Book – Guy with Free Stock Footage by Guy with Free Stock Footage licensed under Creative Commons License CC BY 3.0 [https://www.youtube.com/watch?v=V9HpUPQ9pt0]
  7. Tom Cruise as Iron Man in the MCU – DeepFake Theater by Collider licensed under Creative Commons License CC BY 3.0 [https://www.youtube.com/watch?v=iDM69UEyM3w]

Photos

  1. Hao li portrait by Janek licensed under Creative Commons License CC BY-SA 4.0 [https://commons.wikimedia.org/wiki/File:Hao_li_portrait.png]
  2. Sylvester-Stallone-2014-2 (cropped) by Michael Schilling licensed under Creative Commons Licence CC BY-SA 3.0 [https://commons.wikimedia.org/wiki/File:Sylvester-Stallone-2014-2_(cropped).jpg]

Music:

  1. Come On, Let’s Go (Free Download) [Chillhop/Lo-Fi Beats] by Le Gang is licensed under CC BY 3.0 [https://soundcloud.com/thisislegang/come-on-lets-go]

Leave a comment

Design a site like this with WordPress.com
Get started