I need to vent a little because I’ve hit a production problem I honestly didn’t expect.
I’m around 70% through developing/producing SPEAK FOR ME, and the main character is Nika, a 12-year-old girl.
She is a completely fictional character created with AI.
Nothing about the project is sexual, exploitative, suggestive, or inappropriate. She is dressed normally, the scenes are dramatic and psychological, and the whole film is actually built around protecting a vulnerable child and the reality of child trafficking.
And I want to be very clear about something before anything else:
I completely support strong protections around minors.
Real children absolutely need protection, and I understand why AI companies have to be extremely careful in this area.
My frustration is somewhere else.
The same image-generation system that created Nika consistently throughout the production has now started refusing to generate new shots with her simply because she is a minor.
So now I have a film that is already around 70% built, with an established main character, established visual continuity, existing scenes, and a specific story - and suddenly I may not be able to generate the remaining shots.
That potentially means putting the entire film on pause.
Or worse, putting it on the shelf permanently.
And that’s the part that frustrates me.
If the system had never allowed this from the beginning, I would have designed the project differently.
But when a tool allows you to build a character, generate dozens of scenes, establish continuity, and then changes what it allows halfway through production, the filmmaker is left carrying the consequences.
For AI filmmaking, continuity is already one of the hardest things to maintain.
You spend a huge amount of time locking faces, clothing, environments, proportions, lighting, camera angles, and character identity across shots.
You can’t simply replace the main character 70% into the film and pretend nothing happened.
So this made me think about something larger:
How much can an independent filmmaker safely depend on AI tools when the rules, models, capabilities, or restrictions can change in the middle of production?
Traditional productions have their own risks - actors become unavailable, locations disappear, budgets collapse - but with AI filmmaking, sometimes the “production infrastructure” itself changes overnight.
I’m not criticizing the need for safety rules.
I’m frustrated by the unpredictability and inconsistency of those rules during an active project.
Has anyone else working with AI had a project disrupted because a model, platform, policy, or capability changed halfway through production?
And if so, how are you protecting your projects against that kind of dependency?
"For context, this is Nika, the fictional main character from SPEAK FOR ME."
1 person likes this
Inknomad Films The hard truth is that you're using a retail AI platform, and that means several things beyond this issue: (1) You're subject to any regulatory framework that government or industry imposes, or the service you use wants to get ahead of (such as this issue). (2) Your creative work becomes part of the platform IP, and will be used to generate images for other retail users. (3) The images created for you have already been created for many, many others, even if you haven't seen them before (unless you provided original source material, in which case see #2 above). (4) Copyright provenance on the images and sequences you prompt is specious, and US Copyright Office, followed by other copyright authorities around the world, will only register copyright in specific cases. Even if registrable, the outlook for the commercial value of AI generated video is currently very questionable. The practical solution is, use more traditional tools like Toon Boom for your animation, which are in fact pretty good.
I understand your point, Shadow Dragu-Mihai, and I agree with the bigger issue you’re raising: if you build a production around a retail platform, you’re accepting a dependency on something you don’t control.
That’s actually what this situation has made very clear to me.
Where I’d separate things a little is that ownership, training use, licensing, and commercial rights aren’t identical across every AI platform - each service has its own terms and restrictions. So I try not to treat all of them as one system.
But on the practical filmmaking side, I agree with you: relying entirely on one external platform creates risk. A model can change, a feature can disappear, a policy can tighten, or a workflow that worked yesterday can suddenly stop working.
My frustration here is less about defending AI as a technology and more about the fact that I built roughly 70% of a film around a workflow that was functioning consistently, and now one of the central production capabilities may no longer be available.
Traditional animation tools would certainly give more control and stability, although they also require a very different skill set, workflow, time commitment, and production structure.
So I think the lesson for me is probably not “never use AI,” but "never let one platform become the single point of failure for an entire film."
Inknomad Films The truth is that all AI restrictions can be worked around more easily than people realize, and I’ve explored the limits of what these tools can and cannot generate beyond a simple prompt. That said, I would never encourage using someone’s likeness without permission, especially children or minors, unless it is your own image or you have clear approval from the parent, guardian, or rights holder. If you have original photos, approved images, or your own person, character, or creature, I’d definitely be open to helping. I also shared a few tools recently that can check images for duplicates or visually similar matches if you’d like to use that as an extra safeguard.
Thanks, Dwayne Williams. I appreciate the offer and the point you’re making.
Just to clarify, Nika is a completely fictional AI-generated character - I made that clear from the beginning. And yes, I’m fully aware that when a real minor is involved in any production, permission from a parent or legal guardian is essential.
The point of my post was really a different one: the production risk that comes from depending on a platform whose rules or capabilities can change suddenly.
In my case, I had already built a large part of the film around a workflow that was working consistently. Then that workflow changed, and I didn’t have an immediate backup that could preserve the same character and continuity.
So for me, the bigger lesson is about dependency and production planning rather than trying to work around safety rules.
I do appreciate the offer to help, though, and the duplicate-checking tools sound useful as an extra safeguard.
Just clarifying Inknomad Films, though: even if it's 'fictional,' an AI-generated character is still trained on real people's faces. So be careful with that distinction. The model learns from actual faces and features, so there's always that underlying data provenance to consider, especially with a minor character.
Worth keeping in mind as you move forward with the project.
It's definitely a risk if you have no previous graphic software skills or enough knowledge of the software you're using to adapt to the changes. When a platform shifts its restrictions or capabilities, you need the technical foundation to pivot; otherwise, you're completely dependent on that one workflow staying stable.
2 people like this
Hi Val, I can understand your frustration. When a project is already 70% developed and the character, visual identity, and continuity have been established, a sudden change in the tool’s limitations can create a serious production challenge.
I also think you raise an important question about how filmmakers can manage dependency on AI platforms when their policies and capabilities can change during production. SPEAK FOR ME sounds like a particularly challenging project because Nika’s character is so central to the story. I hope you’re able to find a solution that preserves the visual and emotional continuity you’ve worked so hard to build.
1 person likes this
While this is a sideways comment, I have a quibble with your use of the term “fictional character.” If this movie is a work of fiction, all the characters are fictional!
The distinction I believe you’re drawing is between an AI generated character based on photographs of a real world actor, and an AI generated character with a generated face based on nothing but training data.
I don’t know what to call that, but “fictional” isn’t quite enough.
1 person likes this
I understand what you mean, Dwayne Williams, and data provenance is definitely something worth being aware of.
When I call Nika fictional, I mean that she was not created from, named after, or intentionally based on any identifiable real child. I’m not suggesting that the technology itself exists independently of whatever data was used to develop the underlying model.
And I actually agree with your second point - that’s very close to why I opened this discussion.
The experience has shown me how vulnerable a production becomes when too much of the workflow depends on one platform. If that platform changes and you don’t have another technical route to achieve the same result, suddenly the project itself is at risk.
1 person likes this
Thank you, Karolina Ferguson. That’s exactly the problem I’m facing, Nika isn’t a minor detail I can simply replace or redesign halfway through the film. She is the emotional center of SPEAK FOR ME, so continuity matters a lot.
And yes, this has definitely changed the way I think about production planning with AI. From now on, I’ll be much more careful about having backup workflows before getting too deep into a project.
I’m still hoping I can find a way forward without compromising what the film is supposed to be.
2 people like this
That’s a fair distinction, Mike Boas.
What I mean is that Nika is "not based on any specific real child or actor". Her appearance was generated from scratch within the AI workflow rather than created from photographs or the likeness of an identifiable person.
So yes - “fictional character” describes the story role, but not the production distinction very precisely. “AI-generated original character” or “synthetic character not based on a real person” is probably closer to what I’m trying to say.
Thanks for pointing that out.
1 person likes this
Exactly, Realt Grace. That’s the distinction I was trying to make.
I support strong safety rules, especially where minors are concerned. What caught me off guard was building most of a film around a workflow that was functioning consistently, only to discover that the same workflow could suddenly become unavailable.
It’s definitely a lesson in production dependency. For future projects, I’ll be thinking much more seriously about backup tools and alternate workflows before getting this deep into production.
Thanks! I’m still trying to find a way to finish SPEAK FOR ME without compromising the character or the continuity.
Give it to another AI