Will AI replace Houdini in some tasks in the future? For example, in film and television special effects or visual effects in advertisements. I've been using Houdini for eight years, and it's a fantastic software. I don't think any other similar software can surpass it.

but with the emergence and development of AI, I'm a little worried about our current efforts.After all, AI can achieve about 70% of Houdini's effect with only one-thousandth of the time and money.

You'll never get a neutral answer about AI on the internet. Too many people have skin in the game (on both sides)

happybabywzy
Will AI replace Houdini in some tasks in the future? For example, in film and television special effects or visual effects in advertisements. I've been using Houdini for eight years, and it's a fantastic software. I don't think any other similar software can surpass it.

but with the emergence and development of AI, I'm a little worried about our current efforts.After all, AI can achieve about 70% of Houdini's effect with only one-thousandth of the time and money.

Ai Can't do 70% of Houdini's effect, not even close to that.
where did you get this number?

And people will always want to see visually than use text ,
so think what will happen when sidefx will develop AI within houdini

and don't listen to dooms day people because most of them speak out of self interest.

raincole
You'll never get a neutral answer about AI on the internet. Too many people have skin in the game (on both sides)

Maybe your answer is the neutral one thank you

habernir

happybabywzy
Will AI replace Houdini in some tasks in the future? For example, in film and television special effects or visual effects in advertisements. I've been using Houdini for eight years, and it's a fantastic software. I don't think any other similar software can surpass it.

but with the emergence and development of AI, I'm a little worried about our current efforts.After all, AI can achieve about 70% of Houdini's effect with only one-thousandth of the time and money.

Ai Can't do 70% of Houdini's effect, not even close to that.
where did you get this number?

And people will always want to see visually than use text ,
so think what will happen when sidefx will develop AI within houdini

and don't listen to dooms day people because most of them speak out of self interest.

I'm sorry, I don't have exact number, this is just a guess. I'm a Houdini tercher at a university, and I often see news about the continuous progress of AI, such as the recent Seedance 2.0, which is already very powerful. I also constantly hear people around me saying that what we teach is outdated or no longer competitive, which makes me a little scared.

https://www.youtube.com/watch?v=ScfibDSMXJA [www.youtube.com]
how long time will it takes you to do the same with houdini? from the ocean (with foams+waves>how many Terabytes + computing time?) from rain particles to ship colision + everything?
ahah

Foocus
https://www.youtube.com/watch?v=ScfibDSMXJA [www.youtube.com]
how long time will it takes you to do the same with houdini? from the ocean (with foams+waves>how many Terabytes + computing time?) from rain particles to ship colision + everything?
ahah

first you comparing procedural simulation tool with AI .
AI can't do simulations 
what happen with close-up shot of simulations? it will be correctly?

the second thing , the real limitation its the moment you need control. 
can you make a specific wave break at the exact moment the ship turns? 
can you get consistent results across multiple shots?
can you match it to live action footage?
what happen when director ask for a change?
and thats only part of the problems in AI.
in houdini you control everything and with AI you just hope for the best.

and when AI will be implemented inside houdini/maya/c4d.....think what dcc software can do ,
i think tools like seedance won't be relevant in the future comparing to this combination

but today AI still can't replace dcc software,
i think the future it will be hybrid soltuion,but thats just my opinion.

habernir

Foocus
https://www.youtube.com/watch?v=ScfibDSMXJA [www.youtube.com]
how long time will it takes you to do the same with houdini? from the ocean (with foams+waves>how many Terabytes + computing time?) from rain particles to ship colision + everything?
ahah

first you comparing procedural simulation tool with AI .
AI can't do simulations 
what happen with close-up shot of simulations? it will be correctly?

the second thing , the real limitation its the moment you need control. 
can you make a specific wave break at the exact moment the ship turns? 
can you get consistent results across multiple shots?
can you match it to live action footage?
what happen when director ask for a change?
and thats only part of the problems in AI.
in houdini you control everything and with AI you just hope for the best.

and when AI will be implemented inside houdini/maya/c4d.....think what dcc software can do ,
i think tools like seedance won't be relevant in the future comparing to this combination

but today AI still can't replace dcc software,
i think the future it will be hybrid soltuion,but thats just my opinion.

About the control, it's there, For the breaking wave example.. you could make a grid with some basic rolling waves on it and use it to guide the AI timing. But more importantly, control is overrated, when you tell a director they can have exact control for 20k or pick the best out of these 20 versions that cost less then a few bucks the money is going to win. It's similar to actually shooting a shot in nature, there you have zero control and people can work with that just fine. I'm sure everyone has seen the spaghetti eating benchmark video,. add another year or two and it's a serious production alternative.

And back to simulations, accuracy is a tricky one. it's mathematical models that takes shortcuts with added an rendering step which in itself is also taking shortcuts versus a direct visual generation based on tons of reference videos. Neither is going to be perfect but again, the ability to do it in a fraction of the time and for pennies is going be a driving factor. Add to that how fast AI is improving I'd say it's going to be an interesting ride.

I do think people who understand the dynamics and can actually setup simulations will create better AI content because they know the fundamentals. Similar to how and oldschool photographers who worked with film emulsions and chemical development will be able to make better photographs using a digital camera because they will put more thoughs and effort into it and have good understanding of camera settings.

Fewer people will be willing to learn 3D software like Houdini in the future, as traditional pipelines are gradually being superseded by AI-driven workflows. It’s a disheartening trend.

wanglifu
Fewer people will be willing to learn 3D software like Houdini in the future, as traditional pipelines are gradually being superseded by AI-driven workflows. It’s a disheartening trend.

why? do you think that sidefx can't implement AI inside houdini? or any other dcc software company ?
why people think that AI engine only live on web application? its not.

right now it depent what sidefx will do , Imagine what happen if AI will go inside houdini or other dcc software.

don't forget that CGI is a business like any other business, knowledge/consulting is generally traded for money in this society.
that being said vfx course is closing and you can have all the courses for free now.
https://thevfxschool.com/ [thevfxschool.com]

look like there was no queue to learn Houdini

I don’t think AI will replace Houdini, but it will definitely replace some smaller or faster tasks around it.

For quick concepts, rough effects, or low-budget work, AI will become more common. But when you need real control, revisions, consistency, and production-ready results, Houdini is still much stronger.

So in my opinion, AI will become part of the workflow, not a replacement for Houdini.

When a picture speaks louder than words...
Image from Nano Banana pro. Saturation mask. Squares are not pixels—they're 8x8 pixel blocks. As you can see, the image is assembled from squares at the boundaries, where the gradients of various functions often break. Although the image before correction looks fine at first glance, just move it slightly and all the imperfections become apparent. This is why neural network images look strange and wrinkled to us. This resembles a jagged curvature on a surface, where the surface is formally smooth, but the glare suggests it's not quite smooth and unpleasant to the eye. It should be noted that such defects aren't always so pronounced, and the situation will improve, but for now, at best, they blur these defects rather than create smooth gradients. I'm sure that neural networks will soon learn to make these gradients smooth, too, but for now, that's what we have.

I`ve worked for, let`s say one of the largest corporations on this planet, just a few months ago to promote a new functionality, and a group of guys specialized on generating image and video used the latest generative AI of the same corporation instead of filmed/photographed content.
The amount of content produced to get where the client wanted to get was mind-boggling, and we still didn`t get the best results. Two shots in particular stay etched in my mind: A cat sitting on a laptop and thereby hitting random keys and this was just ONE clip, a few seconds long: The screen showed nonsense, the keys had garbage alphanumerics on them, their shapes were "morphed" between Mac and Windows, the cat`s head movement were robotic and the timing the agency wanted was inverted. I did a helluva work to fix most of it except the odd head movement which we took care of by time stretching the clip and cutting that moving head part out.

Apart from all this, my general observation is, that the images/videos are very very often awfully blurry and full of artefacts, I can`t believe that after so many years of refinement of our tools, from 2D to 3D and the best quality cameras ever built we`re now accepting this kind of low quality crap built upon mostly stolen content. And one thing a lot of people are not getting is: this is the best these generative systems can get - they`ve literally gobbled up whatever they could grab, from low res, heavily compressed garbage to a bit of higher res stuff. And since less and less people will share their content and nobody will produce film/commercial level stock images, there won't be anywhere to go to steal new/better data to feed the beast. It´s also impossible to create "ugly" stuff with the curated general purpose AI`s, just as an ever growing set of filters prohibit various things the current mainstream deems inappropriate.

I also thought about the cultural consequences. We`re now entering an era of less and less new human made creative works. We already are in a cycle of endless rehashing of old ideas, but if wide adoption of AI content happens, everything will be preserved/mummified at a pre-2020 kind of state. Everywhere I go, regardless if Europe or Asia I see the same midjourney-style averaged images of "beautiful" people. It`s an actually depressing mindset the media is trying to push. It`s the superficiality of social media nonsense we already had but amplified by a thousand.