Home Digital Sketchbooks

Erza Draksya — Fantasy Character Design & Illustration

Hi everyone! I'm Erza Draksya, a fantasy character designer and illustrator based in France. I create original characters and worlds using a hybrid workflow (AI-assisted generation + manual Photoshop refinement). I'd love honest feedback on my work — what works, what doesn't, what catches your eye.

Here's one of my latest pieces — Rynasha, daughter of two dragon gods, leaping from a cliff above her father's fossilized heart:"



"You can see more of my work here: https://www.artstation.com/erzadraksya

Any feedback welcome — I'm here to learn and improve. Thanks for looking!

Replies

  • Fabi_G
    Online / Send Message
    Fabi_G godlike master sticky
    Hi! I respect that you're upfront about your use of AI. I suggest you take look around the forum to get an idea of what some positions are on AI use, for example AI Art, Good or Bad?

    If you really want to learn and improve: Ditch the AI part! Looking at the image, I see objective errors which make me question whether you're serious or lack some fundamentals (and as a result can't see what's wrong)?
  • zetheros
    Offline / Send Message
    zetheros insane polycounter
    I wont spend the time to give you feedback if you didn't spend the time to do the work, lol
  • Francois_K
    Offline / Send Message
    Francois_K interpolator
    I am seconding what Fabi_G said.

    Your character has two right legs and one has 6 toes. This is not a small thing to miss especially when your whole generated concept is a character. That speaks volume of how developed your eye is to spot anatomical inaccuracies.

     If your goal as you say is to learn and improve : Start with the basics.

    Also it's rather disingenuous to put the tags "Created with AI" and "NoAI" consecutively on Artstation when you here flat out state you used it.

    Best of luck. 

    Edit: 
    I stand corrected with the "noAI" part. I was not informed properly.
  • Eric Chadwick
    The NoAI tag on ArtStation is not for indicating how the art was made, but to try to prevent scraping.
    https://magazine.artstation.com/2022/12/noli-tag/

    The tag is so poorly designed, because at first glance it looks like a "how this was made" tag, since all  the other tags are setup that way. Pretty unfortunate.

    Anyhow, to the OP, for me the art looks derivative overall, and has some obvious errors as pointed out. This is what we commonly call "AI slop". What Photoshop paint-over work did you do for this? It would help to show a raw "before" shot, next to the "after" result, to help people gauge your skill and provide feedback.
  • ErzaDraksya
    thanks everyone for your return i am ashamed to confess i have missed this issue with feet i will try to correct it :)
  • ErzaDraksya
    Hey everyone, thanks for the feedback on this piece — the anatomy critique on the foot was correct, so I went back and fixed it. I've updated the image in the original post with the corrected version.

    Quick note on my workflow, since it came up in the comments:

    I work with AI generation as a starting point and do extensive post-production in Photoshop — color correction, anatomical fixes, lighting adjustments, texture work, vignetting, grain. This piece in particular went through 12+ iterations on the generation side (prompt engineering for specific scale textures, eye colors, body proportions, costume design) and then several rounds of manual correction for the things AI models reliably get wrong: hands, feet, toe counts, lateral symmetry.

    I'm a character designer and art director by training, not a traditional illustrator — I don't claim to draw from scratch and I don't pretend the generation step doesn't exist. My contribution is in the design decisions, the character lore (Rynasha has a full character bible I've been building for months), the visual direction, and the post-production craft that takes a generic AI output and turns it into a piece that fits a specific universe.

    I understand the skepticism around AI in art communities and I get that 'AI slop' is a real problem when people just publish raw outputs without engagement. That's not what I'm doing here, but I also know that distinction isn't always visible from the outside, and the only way to demonstrate it is to keep working, keep posting, and keep accepting feedback like the ones you gave me — which were useful and resulted in a better piece.

    Thanks again for taking the time to look.
  • Eric Chadwick
    Thanks for the update, nice work on the new foot.

    I can see how ai-gen tools can be attractive for the gee-whiz factor, how much it just goes ahead and fills in all the details automatically.

    But it’s unfortunately a very problematic process, encumbered by deep issues with morality. These models are trained on vast scraped databases of stolen art. They rely on huge energy and water consumption, massive pollution, and strain communities’ local resources. 

    And on top of all that, they produce derivative styles and unoriginal outcomes, despite all your efforts at driving them towards an originally conceived goal. To traditional artists, it just ends up looking cheap. 

    I’m not opposed to new tech and tooling, on the contrary I depend on new developments to extend my craft and expand my career, every day.

    But some tools have deep problems, impossible to ignore the consequences.
Sign In or Register to comment.