Ethics and Transparency
At IMAGIKAZE we combine state-of-the-art artificial intelligence with traditional retouching techniques, art direction, and visual production to create our catalog. We believe that being honest about how we do things — including what technology still can’t solve — is part of our identity, not a risk to it. This page explains, without filters, how we work.
What do we think about AI?
Honestly, we don’t have a closed answer. It’s an open debate, and we acknowledge that.
On one hand, we understand and take seriously the concerns: generative AI does represent a real threat to many creative jobs — photographers, models, retouchers, illustrators — who spent years building their craft through practice, equipment, and investment. It’s a legitimate concern, and we don’t minimize it.
On the other hand, it’s also true what many point out: today anyone can write a prompt and get an image in seconds, with no studio, camera, or years of technical training required. That democratizes access, but it also creates a saturation of content and, for some, devalues the effort behind traditional creative work.
We don’t claim to resolve this contradiction, because it probably doesn’t have a single correct answer. What we can tell you is where we stand: we use AI as a tool within a larger process — art direction, curation, retouching, aesthetic judgment — not as a shortcut to skip creative work. But we don’t claim to be the moral exception in an industry that, as a whole, is still figuring out what this technology means for the future of creative work. It’s a conversation we keep having, not one we’ve already closed.
Commitment to realism (and its current limits)
We aim for the highest possible level of realism in every piece: skin with visible pores, natural reaction to direct light, and real textures and imperfections that keep the result away from a “plastic” or artificial finish. But we’re clear: current technology still has known limitations, and we’d rather tell you before you find out yourself.
Face and skin: the finish can sometimes look less realistic than expected, without the texture or light behavior we’re aiming for.
Hands: although generally well-formed, they may not fully reflect the real weight, tension, or natural movement of a human hand.
Sharpness: we work to avoid blurry or low-definition results, though it’s not always perfect.
On-scene text: text that appears within images (signs, garments, backgrounds) can sometimes be generated incorrectly or illegibly.
Character consistency: the same AI-generated person can show slight variations between images within the same set.
Level of detail: not all generations reach the same degree of definition; some pieces may look more detailed than others within the same collection.
We don’t hide these limitations: they’re part of the creative process we work with today. Our hybrid workflow of retouching, curation, and fine-tuned generation exists precisely to reduce their impact. We’d rather be clear about the current state of the technology than promise a perfection that doesn’t exist yet.
Traceability: SynthID and open provenance standards
Part of our production workflow —images and, increasingly, video— relies on Google’s generation technology, such as Nano Banana Pro and Veo. These tools incorporate SynthID, an invisible digital watermark developed by Google DeepMind that is embedded directly into the pixels of each image or into every frame of a video —not in the file’s metadata— and is designed to withstand compression, cropping, format changes, and common editing. In the case of video, the mark covers the clip frame by frame, so it survives cuts, transcoding, and most post-production processes.
We want to be clear about this: we don’t attempt to evade or remove this kind of identification system. Even if a piece goes through curation, retouching, and professional editing, this mark is built precisely to persist through those processes. We don’t see it as a problem to solve, but as a responsible part of the ecosystem we work in.
Beyond SynthID, the industry is adopting a complementary open standard called C2PA (Content Credentials), backed by a coalition of over 6,000 members —including Adobe, Microsoft, the BBC, Reuters, Sony, Leica, OpenAI, and Google. While SynthID lives in the pixel or frame, C2PA works like a nutrition label attached to the file: who produced it, when, and what tools or editing processes were used, forming a cryptographically signed chain that can be verified step by step. The standard is also expanding to video and live streaming, so we expect to apply it consistently across both formats as support matures in our tools.
We’re just as clear about its limits. C2PA is a solid provenance layer, but not foolproof: most social networks, messaging apps, and email clients strip this metadata when processing or recompressing a file —whether it’s an image or a video— while SynthID, living within the content itself, usually survives that process. Also, a valid C2PA credential certifies origin and editing, not that the content is accurate, properly licensed, or used in the right context. That’s why the file with the full manifest will live in our final delivery to the client and in our own portfolio —the channels where we can guarantee the provenance chain arrives intact— and we’ll offer, upon request, direct verification of the content credential using public industry tools.
We believe in and support this kind of traceability system because it helps prevent the misuse of AI —misinformation, identity impersonation, deepfakes— and helps the digital creative industry grow on ethical foundations. At IMAGIKAZE, technology is a tool for amplification, not a means to hide the origin of our work.
AI-generated characters
The faces and people that appear in our collections —photography and video— are synthetic characters, generated by AI. Our approach when creating them is to avoid recreating or directly referencing the face of any real, identifiable person: they are generated compositions, not representations of anyone in particular. All characters that appear in our work are adults.
We state this directly because we understand that the line between synthetic and real is becoming harder to tell at a glance, and we’d rather you not have to guess. Since these models are trained on enormous volumes of real faces, an accidental resemblance to an identifiable person is a known risk of the technology, not something we can fully rule out through intention alone.
How to report a resemblance
If you or someone you know recognizes an unintentional resemblance in one of our pieces, you can report it directly to reports@imagikaze.com (Report a Concern). To review your case, here’s what we ask for and what we don’t:
What we need:
The link or name of the piece, set, or video where you saw it
What specifically seems similar (face, a particular feature, birthmark, tattoo, etc.)
What’s optional:
A reference photo of the real person, only if you want us to compare directly. It’s not a requirement for your report to be valid — you can describe it in words and we’ll evaluate it just the same.
What we never ask for:
Official IDs or any legal documents
No additional information beyond the above
How we handle your information, no gaps:
If you send a reference photo, we use it only to visually compare it to your report, manually — we do not process it with facial recognition software or any automated system.
We delete the photo once the case is resolved; we don’t keep it indefinitely.
We don’t share your report, your photo, or your email with third parties, or with the AI generation platforms we use in our process.
Sending your report is voluntary. By sending it, you understand that you authorize the review described above, and only that review.
What happens next:
We review the case and reply directly to the email you wrote from; usually within a few days, since we’re an independent project without a dedicated support team.
If we find sufficient grounds for the resemblance, we remove or modify the content.
If we don’t, we’ll explain that too — we don’t leave reports unanswered.
Our contribution to the AI ecosystem
We use these tools actively and consistently —not just as end consumers— and share feedback with developers and platforms, both large companies and open-source projects, always within the terms of use of each; we don’t collect or manage third-party data for that purpose. We keep evaluating new models rather than relying on just one, because that lets us offer the best possible result to those who trust us, and because we believe being an active part of this improvement cycle —rather than just passively consuming the technology— is a form of creative responsibility: the more consistent and realistic these models become, the more it benefits the entire community of independent creators who, like us, rely on them to raise the quality of their work.
What you’re actually buying
When you buy a piece, a set, or a video from IMAGIKAZE, you’re not just paying for “AI-generated content” —anyone can produce that for free in seconds—. You’re paying for art direction, curation, hours of retouching and professional editing, visual consistency throughout a full set or piece, and a final production made for real use, whether editorial, inspirational, or as creative reference.
Precisely because of the technical limitations described above, we don’t promise 100% fidelity or perfect realism in every piece. We’d rather you know this before buying, not after.
An open conversation
This document isn’t a closed text. As technology changes, tools improve, and public debate evolves, this page will evolve with us. If you have doubts, questions, or simply want to push back on something we say here, we’re open to that conversation.