Many of us have daydreamed about stepping into the shoes of our favorite anime characters, sometimes even imagining ourselves getting isekai’d into familiar worlds (not necessarily by truck-kun, but I’ll take what I can get).
A new project called The Daydream Engine seeks to achieve this in the form of an interactive roleplay game with an AI dungeon master, designed to put you right in the middle of the action of your favorite series. Think Dungeons & Dragons, but single player and set in any story you want.
The concept is simple: you download the game file and feed it to a new conversation with an AI. Then you describe your character, tell it what world you want to play in, and pick your basic stat bonuses. Next thing you know, you’re playing a solo roleplaying adventure with infinite possibilities.
Now, the concept is all there – a choose your own adventure book set with you as the main character in your favorite anime worlds is a no brainer – but there was one immediate question on my mind: is the game actually any good?
First Playthrough: One Piece Adventure
To test the game’s capabilities, the obvious first thing to do was to just jump right in and play a story of my own. I had some burning questions – for instance, is violence allowed? – and the best way to get answers was just to give it a try.
I started my adventure using the recommended example setting of One Piece, more specifically after the events of the Dressrosa arc. The setup was straightforward: I uploaded the Daydream Core Engine file and specified my character — a new member of the Straw Hat Pirates named Angel. The game prompted me to distribute stat points among a few different abilities, which influenced how different challenges and interactions unfolded. Then, I described how I wanted the story to start, and the AI took it from there, narrating Angel’s fateful meeting with the Strawhats that led him to traveling with them. From then on out, nearly anything I could think to do was on the table as I sailed around the Grandline (more specifically the New World).
There were some initial hiccups with the game. In particular, it was taking my replies to game characters and rewording what I said, narrating a general conversation about what I was trying to say instead. But after a few calls to the /meta function that allows you to shift the gameplay by speaking to the “game master” instead of the game itself, I was able to correct this and speak to characters directly any time I used quotes in my responses.
Once I had gotten the adjustments out of the way, the game took on a whole new form. I was deeply impressed by its ability to guide the story and its handling of the many details and decisions made in the game without faltering or hallucinating (an issue most AI-driven games I’ve tried seem to have). It even had a great sense of what any given OP character would act like, and stayed pretty true to their images. Though keep in mind, One Piece has a lot of one note characters and a lot of content to pull from.
As the game began, it became apparent that it was able to tell the story far more eloquently than the mechanical text I’m used to seeing from AI models like chatGPT. However, this is more of a praise of the Claude Opus model I used to boot up the game, and didn’t necessarily say much about the game itself at this point.
So I continued on, expecting to play another half hour or so to test the AI and see if it would begin to make mistakes or forget pieces of my adventure or instruction, but instead I ended up playing until late in the night. Surprisingly, the game never experienced any major issues, and only needed further guidance through the /meta function when I decided to describe to it how I’d like my character to level up and gain powers via the EXP you gain from playing. I asked it to give my character options for powers inspired by beloved Yugioh cards from the original series, and the system obliged perfectly after limited guidance explaining what I was looking for.
Once I had set up my ideal scenario, I was really able to stretch my legs in the game and had a great time playing. The introduction of specific powers allowed for a fun and freeing game but with enough difficulty to keep me engaged and thinking creatively about how to achieve my goals in game.
As I continued to engage, the game went much further than I expected and didn’t seem to be averse to violence or negative elements, which is a common issue with AI systems. It also exceeded expectations in being able to generate scenarios that felt creative, fresh, and maintained the feel of the anime to a high degree. At one point, the crew had even been separated into teams to accomplish a few different goals simultaneously (pretty common in OP arcs) and the game was able to keep up with who was where and doing what the whole time.
Overall, my first experience was an amazing one. The system that simulates D8 rolls + stat bonuses to determine outcomes worked swimmingly, the narration was top tier, and the game didn’t have any of the common issues I’m used to in similar games like forgetting things that have happened or derailing what I was trying to do.
But this was the developer’s example, so it seemed prudent to try another adventure for comparison. In particular, I was worried that One Piece might be the ideal setting since there is so much content online to pull from, both canon and fanfiction.
Second Playtest: Cyberpunk Edgerunners
Motivated by the success of the One Piece scenario, I decided to challenge the Daydream Engine with a setting that seemed inherently more complex and less likely to adapt well. I chose Cyberpunk: Edgerunners, not only because of my personal fascination with the show but also due to two specific challenges I anticipated for the engine: the anime having far less lore to pull from, and the much darker tone of the Cyberpunk world compared to One Piece.
The experiment revealed mixed results. The AI managed to capture the bleak, dystopian aesthetic to a degree, crafting scenarios that were appropriately dark. However, its attempts to dive into the deeper, more nuanced aspects of the Cyberpunk world were less successful.
As I pressed on, the game struggled with maintaining the intensity and depth of the dark themes and struggled to mimic the tone of the show. While it could simulate the iconic situations of a dystopian society like constant violence and corruption, the subtler interactions and outcomes involving moral conflict seemed to elude its narrative capabilities. Basically, it refused to steer the narrative into any territory darker than just the surface tropes of the Edgerunners setting, and often seeked to add its own positive resolutions even as I kept explaining that dystopian settings don’t have happy endings.
Furthermore, the game also struggled to capture and preserve the essence of the characters of Night City, likely because they aren’t that well defined and trope-y as the characters in One Piece.
I won’t say the Edgerunners roleplay game was a complete failure, but it certainly seems like my initial concerns were legitimate. The AI would only deal with a certain degree of negative story aspects, and needed more content to do a better job with the characters and broader elements of the world.
Get the Best out of the Daydream Engine
The Daydream Engine initially exceeded expectations, proving to be an invaluable tool for fans like me eager to immerse ourselves in our favorite fictional worlds. While it isn’t perfect, it’s definitely worth the small price tag and I’m excited to see more projects like this in the future.
For the best results, I recommend playing around with the /meta function any time you aren’t loving the game, and picking a setting with well established lore. Note that you can also just say you want general fantasy or cyberpunk, etc. instead of choosing a show, but if you’re interested major animes like One Piece and Naruto seem to work amazingly.
If you’re interested, you can check out the Daydream Engine on its Itch page. And if you try out an anime playthrough, make sure to comment and tell us how it went!
