Clients editing images

Looking for advice here and potentially open up a dialogue…

In the past couple of months I’ve stumbled across 3 separate clients of mine editing renders I’ve created for them with AI. None of them have informed me they’re doing this.

It is in my contract that images should not be edited in any way, but I have never really enforced this. However now, with the ease of AI this seems different. Ideally they shouldn’t be training AI with my images either.

But, what can I realistically do? After all, they’re the owner of - or at least paid for - the work.

I will specify more clearly at the commencement of future jobs that this is a big no-no. But I also don’t know how to enforce it? If it’s in the contract, so what? What repercussions can I follow-up with?

I also don’t want to be difficult as the profession of rendering is already so threatened on various fronts. Ideally, if a client wants to use an image of mine to create an animation or change some objects, it’d be me doing the (paid) work, but if I stipulate this before an agreement is made it’s just another reason for said client to not use my services and go to a cheaper alternate.

I’ve always stood firm that we’re not going to lose our jobs to AI (yet), but today I’m feeling a bit of impending doom…

(My only thought is to stop doing ‘per-image’ fee work and offer more packaged deals; €x amount for a range of images, moods, animations etc in the hope this will secure my position a bit…)

Any advice much appreciated!

Thats interesting, I’m curious what they did with those images, object removal/replace, weather/lighting changes etc?

Thats a tricky one I suppose. There is a distinction between training AI and using it. What I mean by this is that if you use offline AI models in something like ComfyUI then that AI model is technically not being trained by the images, its brain is ‘frozen in time’ so to speak. Its also private and not sent back to a server. If the client was then to add the images to the AI’s learning data then that would be training the AI.

It’s tricky like you said since they paid for the images so technically if the AI was not being trained then the distinction is important, a loop-hole consideration. Just some things to consider when you are re-writing your contracts.

With things like wan-video generator Im sure it wont be long until a client can make a 30 second animation from 1 or 2 preview renders.

Thanks for the reply James.

One client took my exterior renders and ran it through AI to make them winter scenes. Another 2 clients used my images to generate video-clips from the still image. That’s all lost work (aka €) for me.

Feeling a bit dismayed all of a sudden… just need to up my game, I guess, in terms of work + artistry, client relationships and professionalism - not sure there is much else that can fend of this threat.

Ha, yeah I followed a tutorial on youtube about this using control-net and depth estimation techniques. Attached images (not my work/tests, this was examples provided by the youtube presenter).

Just a matter of time I suppose. Looks like nearly anyone without 3D knowledge will be able to compete in some capacity soon enough. They just dont have as much control as the 3D artist so currently would probably be quite a frustrating experience for both ‘artist’ and client eg. when they ask for something to change and it changes more than it should.

Here is some ways of using those methods I mentioned for architecture (There’s also a few talented 3D artists showing how they use this in their workflow in other videos):


Hope you dont mind me hijacking your thread Nicolas, just thought you might also want to take a look at this last one a 3dsmax → ai workflow.

Hey, not hijacking at all! Thanks for sharing James. This is strangely reminiscent of a tool + workflow I’ve recently built myself (at a much cheaper cost, I’m sure). Side note; but this VizAcademy guy has always given me weird vibes - sorry if he happens to be reading this!

A. Good to know I’m on the right track in terms of engaging with AI
B. The comments section on YouTube is pretty brutal - has restored a lot of faith in my instincts/hope that AI is indeed not the threat I’ve recently become afraid of.

That brutally sucks.

It might be time to state this very up-front when starting work with client, it is copyright infringement after all, you’re only granting the usage of work, not transferring any rights. But yes, enforcing this when you’re not DBOX, isn’t exactly foolproof and easy..

Also, perhaps time to see if some of the pixel-level tech protections even work.

Me, I am becoming wood-worker sooner rather than later..

Insert one of the promo videos here of some humanoid robots company that are popping right left and center lately…

Ha yeah I was reading through the comments - it is brutal! While I’m always interested in these tech advancements we know how picky clients are with materials/colors/dimensions/details and essentially nailing the brief. Might take a few more years until these tools get to that point.

On that though I really do like the way D5 are going with their ai tools (would be great to see some of this come to Corona). Such as ai scattering tools for gardens/plants with real geometry, lighting for different times of the day, auto pbr material by reference selection, basically empowering the 3D artist to focus on the real stuff and leaving the laborious tasks to the ai. Although with the pbr stuff ive been seeing lots of generated wood lately that looks terrible!

Good point, suppose that also applies to providing clients ai ‘assisted’ work too. Not to mention giving ai NDA data is obviously a breach of conduct (drawings/plans/client IP).

D5 became just data collecting scam with all those new tools. Very convenient but whatever you create will pretty much by default end up on their servers in China.
Same issues with sending sensitive material to Nanobanana or others I guess. All such work should only be done locally if you ask me - but it’s possible nobody cares anymore.

ps - carpentry as exit strategy also crossed my mind. or tiny photo minilab shop..

Know to “Glaze” your works.
https://glaze.cs.uchicago.edu/index.html

Stop using the word “AI”. It is marketing strategy.
Are ‘smart phones’ getting smarter? Or is production/providing services getting automated better?
Don’t be the rat in the race (“Rat City”). Time is to slow down and think. Machines can’t :wink:
This process is more of a ‘machine learning’, somewhat self-correcting process. Without much of real thinking, but satisfying the norm - integrated award system. And it’s huge fast slop in large quantities. Starting to blow back on many fronts. Especially in and via the ‘psychological’ realm.

Differentiate yourself, your mastery, a skill set to that of the machine:
Start by understanding the difference between Hallucination and Illustration - VS - Predictive simulation and Visual prototype.

And when your contract gets breached: “Fuck you, Pay Me!”

Hah yeah Im obsessed with watching off grid home building. Theres a a really great one Im into atm on youtube called - Bushcraft Farm. Very talented young man with great skills (and a very impressive forest home/garden).

Agree, I only use local/offline ai models these days. Privacy/data protection should be number 1 priority in my opinion regardless of what you are doing.

Must be something about us sitting down behind a desk all day! I also dream of wood working… or basically any physically engaging craft.

Well damn, the encouragement I got from this thread was short-lived! If the top-tier guy is planning his exit, what chance does a mid-tier professional like myself have!

Carpentry in wholesome is tricky and also F-ed up (aka devalued) by the accelerated/automated machined work.
In this regard, my specialty is wood carving and custom made pieces from selected, properly treated wood. Did it before my CGI practice. And is now one of many deficit professions. :wink:

Glad you brought up the subject, interesting conversation so far :smiley: Also your work is great Nicolas, I wouldn’t call you mid-tier.

Funny you say this, I was a carpenter (junior) by trade before I switched to 3D back in 2000’s. Im actually terrible with my hands and make lots of blunders, but in Juraj’s home land, carpentry is still a valuable skill and I am also looking to build a cottage out there in the Orava area at some point, so I hope to sharpen those skills. Also having a CNC machine and some other new tools can also be invaluable for the foreseeable future (well atleast for the next 10-20 years). In the Orava region you still see people on horse and cart from time to time :smiley:

While my Veronika is already doing some wood-work (mainly CNC camping stuff in cars), I would just cut my fingers off :- )

And also, not sure who is gonna buy all the custom and expensive carpentry if we all lose jobs to AI.

I am also sort of semi-serious, I honestly just don’t know, I am neither catastrophizing but neither do I really feel much optimistic… while I know some will use it to maximize productivity even in visualization (those probably already do, thankfully I didn’t log into LinkedIn hellscape in two years), I don’t think I will be one of them.
I don’t even really find it fun or enjoyable that much..

This winter I already had my first project where the client used AI to create the full visual aspect of brief. It was much to my chagrin..