I’ve got a few concerns with it at the moment.
1) The writing
Some of the dialogue feels too long and a bit unnatural. Characters often explain everything at once, and in places repeat the same point. The intro could be shortened, with more of the backstory revealed naturally as the game goes on. it should have more interaction, you could easily make the intro a playable part of the game, have the fight something you could interact with with different outcomes, etc. Show/Interact before tell - even with AI (and especially with AI) you could remove long "text" scenes, and instead have more interaction and it is a lot easier using AI for vibe coding/text/etc etc.
2) The graphics
I think the game needs a clearer art style. Some images are artistic realistic, others are anime, and the NPCs look different again. The more realistic images also have that overly detailed AI look, which can be distracting.
Either style could work, but it would look much better if the character art, NPCs and backgrounds all felt like they belonged in the same game. Which mostly leads into the next.
3) Consistency
The characters and rooms change between images. Doors move, beds shift position, and objects appear or disappear. I know this is a common AI issue, but it makes scenes feel less believable.
It may be worth using ComfyUI, LoRAs or stronger reference images to keep characters and locations more consistent. It may be worth looking at online tools such as Leonardo AI, Ideogram, OpenArt or Scenario, which offer character-reference or custom-model features without requiring a powerful graphics card. Cloud versions of ComfyUI are also available if you need more control. Otherwise do what some others do, use ComfyUI logic, and build your own Custom LORAs for characters, locations, rooms, etc.
There is clearly a lot of work going into it, but tightening the writing and making the visuals more consistent would improve it a lot.
-------------------------------
Because of my own creative position, I am not really in favour of AI-generated graphics. That said, I know people are going to use them anyway. The cat is already out of the bag.
What matters is how they are used. If you are going to use AI, you need to avoid falling into the trap of producing generic “AI slop”. The goal should still be to create something with personality, a clear visual identity, and your own creative direction. This includes not just the graphics, but the writing style, the UI, the fonts, the way it appears. Make it unique, make it you.
The tools should support your vision, not decide it for you. If the limitations of the current AI tools are shaping the characters, environments and style more than the creator is, then the end result is going to feel generic rather than personal.