|
Especialy since we now get dialogue in AI, and we had some progress in accessible scene consistency. Some people are proclaiming the death of traditional media production, while others go in full tactical defense mode defending traditional production. So, from my relaxed outside point of view, I thought I make a reality check, what I think has to be said about current AI. Also I think the discussion has lost a bit of track. And is a bit overhyped in emotions - Spoiler, unlike the first article this will be much more optimistic This illustration tries to convey my feelings:  The first interpretation of the illustration: Your job is save - for now Besides still having quality problems (Resolution, clip length and consitency) AI still needs heavy user intervention (Read the first article I wrote when the AI hype barely started, on how I think AI will kill your job. I still think it is valid today,) Lets rundown on some AI usage cases today without the hype: Yes you can use ChatGPT to write code But if you can't code you will have a hard time to write complex programs. Yes you can get shiny concept art out of midjourney But without a proper eductation / training and experience, like working with clients you will have a hard time elevating it to something you can actuallly use. Yes the mesh generators are impressive But without proper modeling / sculpting experience the result are more or less useless. Yes you can make films (sort of) But without proper training the results will be boring and lifeless. AI can't coherently edit itself with proper pacing, and will not in the next years until all other coherency problems are finaly solved. And even then people with filmaking experience will have it easier to create competent films. Yes you can do presenter videos But lets be hoonest these are even more lifeless than real presenter videos. Human intervention can do wonders, especially when actual footage is used. Yes you can write an essay with chatgpt But it will be pretty generic if you don't elevate it with your own style. Yes you can do music with udio But it sounds far better if you professionally edit/mix the output. The stock footage market was already dead thanks to more accessible technology.Sorry. Yes you can make Illustrations for articles For print you are running into a resolution problem for online...yes, but to have a good idea and refine the output you still need someone competent In short the actual craft has not replaced, but imho shifted. And ironically more towards the creative side with all of its decissions. Sure if you do some youtube shorts / ticktok clips / memes, most of the work can be done automatically, but even THEN your decission to edit it further, or keep it is a vital part in the creation process. If you don't filter out the visual displeasing content, it does not matter how good your AI is, ...iin the end you do a creative decission. At the moment the AI hype reminds me a bit of using Tarot or IGing; The real magic is not laying down the cards / the description of them, the real magic happens in your head interpretating the result. If something fits yor situation you do the confirmation bias game, if something fails you ignore it. Thats imho also the trick why you get 4 images in most image generators at the same time: If only one gets close to your what you want AI did wonders. The 3 others wich are below your expectation will be ignored. In short, what I'm saying is that AI can create material to work with but human intervention is still able to elevate it, and this requires traditional craftmanship, experience and taste. And I think this will be a key factor for a long time. The second interpretation of the illlustration It fails in progressive iterations The actual usage as a tool in a larger production process is meh. Yes its shiny, but its integration into a rpodcution cylce is less optimal. Sure you got some tools, like inpaint, masking, depth generation etc, but for the actual workflow with these tools you are still in a "accept or die" situation. You can't refine the AI output that easily. The closest are llms wich allow you to correct them, but for all other, especially in the visual field refinement usually means hope for the best and add additional manual labor. So for professionals it currently means, that they have to charge for the worst possible outcome. (AI fails). BTW sidenote for all 3d graphic enthusiasts out there, there was an interesting talk (or a section of the talk) on the blender conference, where a guy feeds is render result into an image generator, and uses its output to improve his own renderings. ... I like this thinking outside the box. And its a usage scenario where AI is used to improve the skills. The next section is my personal thought, not a professional POV, in short I'm defintively out of my turf on this one but since everyone seems elligble to write on this topic, I don't have that much remorse to add this to the article. Take it as a grain of salt: AI will be the death of Hollywood ?!111!! No, I think we are very close to use AI in previz to frankenstein together a watcheable preview of the movie....and this will save film production creativity wise. This WIP movie will allow filmakers to communicate their vision of the movie better to themself, to studios, to investors, and across the whole movie pipeline. New concepts and ideas are hard to communicate, especially if lot of money is involved, and invesors rather resolve to proven concepts. An lower quality AI version can, to a certain degree, open up hollywood to new ideas..if a producer can actually proove that a new movie "works" before the production starts, this might lead to more original ideas. Don't get me wrong, I believe the final version of the movie is done the traditional way to avoid copyright problems, and have a coherent movie, But since there is a good reference, all departments involved can SAVE money eliminating a lot of revisions and time. Idealy, You don't even need to be experimental on shooting days on set (requiring smaller sets), and you don't need extensive vfx wich is prepared for all eventualities, since you did all the experiments in previz. Heck even an AI result the director / producer does not like is a good reference to communicate the idea better and more cost effective. And no this is NOT the death of concept art and previz, on the contreary with the help of AI and with an output wich is closer to the final film than ever before, becomming more and more a exact replica of the vision / end result Those departments will gain in importance and more time and money will be invested. Maybe in 10 years the whole department will become mostly AI operated but in the meantime those will have more work polishing AI results and integrating it with their own work. EDIT: Of course this all depends on how cheap and how fast (with the all needed iterations) such an AI supported Previz will actually be. How much time and money does it take where 60-80% of a 90 Minute preview is AI generated. Thats the elephant in the room. But hey again.....I'm outside of my element, so don't blame me if this does not happen. And besides that, I never won significant money in the lottery...if this has any relevance about me prediciting the future :) If independent content creators can challenge hollywood in the near future is a topic I don't want to do any prediction, its imho still to soon. Well that sums my current thought on AI hope you could get something out of this ramblings. Defintive gets a follow up. Sidenote: The idea for the banner image suddenly popped out in my head. So it is quite likely I have seen it in the past, if so send me a mail. Banner done traditionally in 3d and illustration in krita / photoshop. No AI used. sidenote 2: Don't start with the "in 5 years" argument, even people who create AI tools don't know what will happen in 5 years. |
