Developing a Way to do Action Scenes Without Money

  Blender Dumbass
August 29, 2025

👁 46

https://blenderdumbass.org/articles : 👁 1
https://feddit.org/ : 👁 1
https://px.madiator.com/ : 👁 1
https://phtn.app/ : 👁 1
https://lemmygrad.ml/ : 👁 1

#vfx #cgi #cars #blender3d #blender #b3d #movies #filmmaking

License:
Creative Commons Attribution Share-Alike

I'm in the middle of developing a movie in which there is a car chase. You know me I can't live without a car chase. But this time I'm trying to make the movie properly. By "infiltrating the movie industry" as some people suggested to call it. Which means ( since I live in Israel ) I need to write something that is possible to do in Israel.

You can make a descent action scene without much money. Michael Bay sustained a feature length worth of chase sequences in Ambulance on just $40 million. Yet the reality of the situation for me is very grim. "Ambulance" is made in USA. A country with almost half a billion people domestically, and a movie industry that is operating world-wide. Israel has only about 10 million people in it, third of which don't really know Hebrew all that well ( which means they don't go to cinema ). So even the best films done here barely scratch 2 million in box office. Getting $40 million to shoot a first feature film in Israel is not something that I can expect to happen. So instead I need to come up with something else.




Thankfully I know Blender.

After watching Shazam! 2 with extremely bad CGI cars and then watching Havoc that has a similar problem, for a second there I thought I was stupid. Back in 2018-2019 I did a test movie to see if I can make a car chase on 0 budget. And yes, I can. It just looks like shit. In the whole film, only about 2 and half shots of it are convincing. The rest looks like terrible VFX done by a student beginner.




And yet those two movies sparked in me some kind of competitive spirit. Shazam!'s cars were rendered badly. I knew what could be done to improve that. Havoc's cars looked amazing, but they had cartoony animation. I knew what could be done with that too. If you combine the good parts from both films into one scene ( which in theory I can do ) maybe it can actually work.

I could do a James Cameron approach to Visual Effects. I could research the hell out of the visual aspect. Focus as intensely as possible on every little detail. And for the movement, I can pretend that I do it for real. For the "Avatar" movies, to escape the fakeness, Cameron used a very smart approach to film-making. Most directors given the freedom of complete control when it comes to CGI shots, would go crazy and stage shots that are obviously impossible in reality. Therefore giving away the fact that it is fake. This is what happened to Gareth Evans with Havoc. Instead I can direct everything the same exact way I would if I had the money to do it for real. Which means, with good enough visuals, I could produce shots that would work in edit with real footage without it being jarring. And therefore selling the VFX.

So I went outside took this photograph...




... and decided to slap a Neonspeedster from Moria's Race onto it.




After being pleasantly surprised with Blender's improvements when it comes to denoising and shadow catchers ( which were a pain in the ass in 2018 ) I quickly realized that it doesn't work. The shot looks very fake indeed.

This was due to a few things. First of all I am a stupid idiot and I didn't use a proper HDRI background for it. In the best case scenario I should have captured a 360 panorama of the place in HDR to use for the light. But with the camera I have available to me this is not possible really. So I used a random sky HDRI file I had on hand. And therefor non of the stuff in the scene is reflected in the car.

Then there is the fact that the car was made for a cartoon. The model is too smooth. The edges are too sharp. Things that would not happen in reality.

So I tried again. I found a good HDRI image on Polyheaven ( not perfect but much better than generic sky ). And I changed the car itself to be more realistic. And here is the result.




Still not perfect, but already one million times better. Actually there was another shot that I think turned out even better.




So I knew what I need to do next. I need to go out there, photograph some car that really exists and try my best to copy it as much as I can. That would give me the best chances at making this possible. So I did, and I found this decrepit looking Honda Civic right under my house. It was perfect.




Look at the texture, look at the various things selling it as a real used piece of equipment. Something that has a good potential to blend-in with the rest of the shot perfectly without you noticing.

For 3 and half days I modeled this stupid car ( I was going to work and taking breaks, and most of the time I didn't have caffeine in the house ). And you know what the result looks like...




All I see is flaws. I see how not quite there the compositing of the car is. I see how not quite there the copying aspect of it is. It should be more matte at the top. And the colors are not quite there. The windows on the original are dirtier. Not much, but there is a visible layer of dust on them. I also cheated with the lights, by using the original image there instead of properly modeling them. Also the original has some out of shape parts that I didn't do. And some things are either missing or out of shape on the replica.

But that said. This render is already much better looking than anything I did in that test of mine in 2018. So I think there is a very good potential if I go through everything and make science out of it, I could make something rather cool here. Perhaps I might need another little test. Not a whole film. Just a few shots of cars like this ( maybe this one ) driving.

In any case. Even though I'm slightly disappointed, I really like what I can do. And I think it is hope that the project will be good.

Happy Hacking!!!

JSON Markdown

Help

Subscribe RSS

Author
Website

Share on Mastodon

Share

Help

You can comment from Mastodon.

Go

Post