There's a post on Reddit that sums up something important about where AI generation is right now.
The user spent two hours figuring out how to bypass Seedance 2.0's face detection. His method: split the target face into a 9-grid portrait, convert it to line art, and feed it through the model. The face detection was fooled. The video generated successfully.
The post got hundreds of upvotes on the Seedance subreddit.
This isn't an isolated case.
What Real Users Are Actually Doing
Seedance 2.0 — ByteDance's AI video generation model launched in 2026 — quickly became a mainstream choice for AI video creators because of its quality and low price. But it has one problem: face detection.
Upload any image containing a face — your own photo, a celebrity's image, even a cartoon character — and the system rejects it outright with "Face detected." No appeal. No manual review. No exceptions.
So creators started finding workarounds.
9-grid overlay, cropping, downscaling, style transfer preprocessing — all kinds of "hacking tutorials" spread across Reddit, Twitter, and YouTube. Creators would rather spend extra time preprocessing images than give up on using Seedance.
Why?
Because faces are one of the core building blocks of creation. Character design needs faces. Storytelling needs faces. Emotional expression needs faces. An AI generation platform that blocks faces is essentially cutting creators off at the knees.
The Real Problem Isn't "Bypassing" — It's Unmet Needs
When AI companies market their video generation tools, they emphasize quality, speed, and cost.
But nobody asks users: what do you actually want to do with this?
A creator who wants to make a roleplay video needs to put their favorite character's likeness into it. A user who wants a memorial video needs to turn a photo of themselves and their family into motion pictures. An independent creator wants to use AI to make a short story film with a realistic-looking original character.
None of this is illegal. None of it infringes on anyone's rights. None of it harms anyone. But Seedance doesn't care. Its face detection is automated, indiscriminate, and one-size-fits-all.
When creators bypass the detection, they're really just fighting for creative freedom in their own way.
Where the Line Between Copyright and Creativity Gets Blurry
Of course, there's an unavoidable problem here: face detection exists for a real reason — preventing deepfakes and protecting portrait rights.
In 2025 alone, celebrity impersonations by AI increased 81% year-over-year. Deepfake pornography, fake news, fraudulent ads — these are real harms that exist.
AI companies have every reason to defend their face detection.
But here's the issue: a blanket filter doesn't just block "bad actors" — it also blocks ordinary users who just want to create normally.
A person using a photo of their own grandmother to generate a birthday greeting video shouldn't be flagged as a "deepfake risk." An independent creator using an original character design they made themselves shouldn't be auto-rejected by the system.
The logic of filtering is: if we can't distinguish precisely, we'll treat everyone as a potential risk. This logic protects the platform, but it sacrifices user experience and creative space.
Creative Freedom Needs a Smarter Solution
This is the problem Moonlight has been thinking about.
AI's value is in unlocking creativity — not in trimming it down with excessive restrictions.
When mainstream platforms use increasingly strict content filtering to "protect" users, we're choosing a different path: we trust our users, we trust creativity, and we trust that people make valuable things when they're not unnecessarily constrained.
This doesn't mean there are no limits. Moonlight's rule is simple: legal, authentic, do no harm to others.
Within that boundary, creators should have the maximum freedom to explore.
We're building Moonlight's image generation capability now. Like our conversation product, it will be a creation tool without unnecessary filters. You can use it for creative design, character building, artistic expression — without worrying that your work will be mistakenly flagged by an overcautious system.
This isn't about circumventing regulation. It's about exploring a smarter balance point.
When Users' "Hacks" Reveal Platform Failures
Back to that Reddit poster who spent two hours working around face detection.
The last line of their post said:
"I'm not trying to do anything wrong. I just want to create."
This sentence deserves serious reflection from every AI platform.
When users would rather spend extra time, use every trick available, and jump through hoops to complete their creative work — it means the existing tools are failing them somewhere.
It's time to ask: who should really be drawing the line for AI generation tools?
If you're tired of hitting walls with AI generation tools, try Moonlight — a platform built for creators, not for filters.
