Following on from the great tutorial technique by Halim Alrasihi yesterday I thought I would use his superb technique to give an idea how much work goes into A.I video. At the end I give a rundown on gens, costs and time taken to produce 2 minutes 30 seconds... of videoshow more

So the conclusion is; We just push a button Remember the 20 minute bond parody I did (Pre-Nano Banana Pro, No storyboards) and all the time inpainting X creators into the video? I never even kept track how much that cost but judging by the figures here it was way too much

For anyone that wants to do something similar, the ratio here for the entire thing is aspect ratio 8:9 Boards on top and video underneath in which case it needs both shrinking in the editor to place one on top of the other Thanks to @HalimAlrasihi for the aspect ratio size🔥

Well done Francolli! Combining this workflow with Keyframes is very smart. I also really like the music and atmosphere. Thank you for sharing your awesome work!

Its a great technique Halim, thanks for your post showing it in its glory yesterday

@HalimAlrasihi Yes it will only keep getting better from here 🔥

@HalimAlrasihi Thanks Emily. I wanted the prompts as the video and the storyboards are moving at the same time. i tried moving the storyboard and video on the left of the screen and scrolling prompts on the right but the prompts are so long they would need to scroll really fast

@HalimAlrasihi My honest Reaction

@HalimAlrasihi A picture of pick up a pencil generated entirely with an a.i art package? Who's art was that stolen from?

Just like to say thanks to everyone who have shown some love to this concept and the movie😊 The Patreon page for the movie is now live, so anyone want to call by I'll get the coffee on and maybe we can discuss shots, scripts, story development etc

@HalimAlrasihi Thank you for this. You did amazing work! I'm using AI to help flesh out my Toshiro story and see if I can have it made into an epic anime series.... who knows, maybe it can all be done with AI and I can skip the big gatekeepers of the entertainment industry!

@HalimAlrasihi Thats the way to think, avoid outside distraction and just go for it. Always have the minset that you dont care what anyone else thinks, do it for you. If you like it then its job done, let the chips fall wherever they like

@HalimAlrasihi I do want to make it my full time job and as a 55yo man who never did focus on my art like I should have, I never became a full time artist and got stuck in manual labor all my life. I need this to be a viable income as well...

@HalimAlrasihi There will always be a strong market for Anime/manga and that whole school of animation. Personally I'm not in that world, its an entire branch of animation that passed me by, but I do see the attraction

@HalimAlrasihi Looks like shit

@HalimAlrasihi The end result is great though. I think this year it will become obvious that performance capture like Kling motion + Sam3 based composition is the way to go for acting. Because bad acting > AI acting

@HalimAlrasihi I agree, even my actiong through mocap and speech to speech will add a lot. Thats my plan

@HalimAlrasihi Great breakdown. The time + cost perspective really shows how much work goes into high-quality AI video.

@HalimAlrasihi Yep, way beyond 10 sec clips on X, but full productions are the natural path where this is all heading

@HalimAlrasihi 42 hours for only this? And so expensive.

@HalimAlrasihi Without deep prompts, which take the bulk of the time, you get bad gens which cost more. So the more time you spend with zero cost writing the less the production costs. Mistakes cost money, the art is dont make mistakes

@HalimAlrasihi I know. I made a 22 minute video. Not with this quality, only by Grok. But it took me like 60 hours of work or so. Cost me like 30$

@HalimAlrasihi It's a start, but also proof that AI has a long way to go to produce consistent and believable outputs when it comes to filmmaking. It's still too much Uncanny Valley. Wooden humans, horses with arrhythmic footfall patterns, wonky physics and fluid dynamics, etc.

@HalimAlrasihi Yeah, I get that, we aren't there yet, both you and I would obsess over those details, I'm in your camp, I won't be happy until all that stuff is ironed out, but the caveat being that I sometimes obsess too much on detail and don't pay attention to 'how do others see it'

@HalimAlrasihi A lot of the issues are inherent model problems that can't be properly mitigated by tuning the prompts. The model knows how a human or a horse looks, but it doesn't know how they "work", so it produces an approximation, not a reproduction. Newer, bigger models will be interesting

@HalimAlrasihi The only problem I see repeating again and again is that the start and end frames always have easy-in/easy-out between shots. There’s no linear motion in every shot, and that’s what makes it noticeable. I hope AI models can fix that in the future.

@HalimAlrasihi I agree, in blender you can adjust that, with a.i i'm sure there is a prompt way around that but i havent found it yet, gens seem to want to start static and speed up as they go, full motion from frame 1 onward is tough

@HalimAlrasihi I sometimes get around the linear motion issue (easing in-speeding up-easing out) by adding into the prompt right at the start, "constant even motion from frame 1." Although from tests its still hit and miss

@HalimAlrasihi Awwwww my goodness so far so good 😭😭😭😭😭😭😭😭😭

@HalimAlrasihi Love this. People massively underestimate how much work goes into “AI video” it’s not one prompt, it’s direction + iteration + cleanup. A breakdown of gens + time + cost for a full 2:30 is super valuable, especially for setting realistic expectations. Excited to see the numbers.

@HalimAlrasihi Oh that's cool to watch it play like that!

@HalimAlrasihi I wanted to have this in 16:9 with the video on the left and the prompts scrolling up on the right but my brain isn't big enough to make that work, lol

@HalimAlrasihi Great post. No matter how much I try to just push a button and get a final product, I agree—pretty much 100% of what I end up making is already visualized in my head before it's generated.

@HalimAlrasihi this looks fucking awful.

@HalimAlrasihi Amazing Sir! Many thanks for showing this! (Bookmarked for my future efforts as well) 🙏

@HalimAlrasihi Congrats took you no time to use someone elses art in AI

@HalimAlrasihi Nice one mate looks fucking horrible

@HalimAlrasihi Awesome breakdown. It also illustrates well the amount of planning you do before generating the images.

@HalimAlrasihi Everything in this video stuff is 90% planning 10% actually animating Laborious stuff

@HalimAlrasihi I want to see the end of this movie

@HalimAlrasihi So do I Any idea where they are going?

@HalimAlrasihi Good job.

@HalimAlrasihi This is mindblowing

@HalimAlrasihi Its even more mindblowing with the 16:9 aspect ratio version with all the stuff in this one placed to the left and the prompts scrolling in real time in the panel on the right. Information overload, kinda gate me a headache watching it. Like a HUD in a jet fighter, lol

@HalimAlrasihi Great and smart moves

@HalimAlrasihi Thanks for the breakdown.

@HalimAlrasihi It's getting there, the problem still lies in the mechanical dialogue and then transitioning from one frame to another, there's that second delay that breaks continuity. Good job in consistency, dynamic shots and tracking, overall feel. Can't wait for dialogue to catch up.

@HalimAlrasihi Dialogue can also be prompted, person A says this, person B says that, but t can be prompted way better. 50 ways of saying hello. How are you today, from friendly to sneering to downright unhinged and threatening. All that can be prompted but not many do.

@HalimAlrasihi The actual dialogue was great, pacing, writing, etc. The delivery of the AI characters still feel like there's a beat between the interaction and flow of natural dialogue. When people talk, their words spill over one-another, and there's subtle inflections in delivery.

@HalimAlrasihi This is why i went for gothic and formal, the language delivery is cleaner, menacing is easier to prompt A pack of high school teenagers would talk over each other, laugh at each other, mix up their words, be stoned or act weird That would be a real challenge to prompt

@HalimAlrasihi Great work all in all. I'm just nitpicking but definitely great fidelity. Bravo!

@HalimAlrasihi Dont fret, criticism is good, it drives improvement😀

@HalimAlrasihi This is a great way to showcase what you did!… most people really have no clue how much effort goes into ai filmmaking.

@HalimAlrasihi tell me about the music!

@HalimAlrasihi The music is a soundtrack I have had sitting around for a while via Udio. I have a lot of these I created with long projects like rhis in mind. This one was 20 minutes long before rhe movie was created, but went back and re-edited it to tailor it for the 2 mins of footage

@HalimAlrasihi Spectacular, my friend! 🙌

@HalimAlrasihi Awesome ~~

@HalimAlrasihi Nice breakdown, the time and cost numbers really highlight the workload

@HalimAlrasihi How Do you generate a precise and long prompt? Can youngive an expample?

@HalimAlrasihi I did a prompt to generate the original image of the coach and horses, then another long JSON style prompt using that image as an input to create a storyboard to provide 9 images to start the story. Both are on my patreon page so go check it out, everything is free over there