Video wird geladen...
Video konnte nicht geladen werden
Opus 5.5 can now turn a single character image into a playable 8-direction RPG character through MCP. AutoSprite has been in development for almost a year, and the workflow is getting pretty interesting for AI game dev. Give it one character and it can generate: > 8-direction isometric sprites... show more
26,061 Aufrufe • vor 11 Tagen •via X (Twitter)
31 Kommentare

try AutoSprite: it also has an MCP endpoint for Claude Code + Codex:

want to see what you can build with Claude Opus 5.5? QuadcodeAI gives you access to the latest Claude models and other top AI models in one place. get 1,000 free credits and start building:

Assets through MCP, code through MCP. The missing piece is the last mile: the same agent shipping the build and reading how real players behave

8-direction sprites from one image is the step that used to eat days. How consistent does the character stay across angles, especially the back view where the source image has no information? That's usually where generated sprite sheets fall apart.

one character image to all 8 directions is crazy…almost a year of work really shows

单图变 8 向 RPG,你这套工作流复用成本多高

Equipment across directions is an interesting part of this. Can you change a weapon or outfit later without regenerating all the base animations? That would make iteration a lot easier.

one image to 8 walking directions is the boring part nobody wants to draw

Sprite sheets killed every game jam I ever tried, so this is really cool.

Having the coding agent call the asset tool over MCP and use the results while it builds is the real shift. How do you keep the 8 directions consistent once equipment changes? Do you regenerate the whole sheet per item or layer them? More Three.js + AI:

How much further is the quality improved utilizing auto sprite, vs have opus make the spirites themselves via code?

How can I recreate this ?

Doing the walk cycles for each character used to be a heavy task. I actually tried doing this for Playground characters as well before switching to using a rigging! Would love to see examples where push this to the edge with complex characters!

is it better than pixellab? i am struggling trying to archieve good results with it

can you change one animation without regenerating the whole sheet? that's the bit i'd want once the character is already in a game.

No hate to you directly, but I'm so tired of all these projects being monthly recurring payments. Whatever happened to just releasing shit to have some fun

Consistency across all 8 directions is the part that usually breaks, so letting the coding agent pull sprites over MCP is the real win. Curious how it handles equipment swaps mid-animation. That's where most sprite pipelines fall apart.

ok this is actually cool. does it keep the character consistent across all 8 directions or does it drift?

Keeping one character consistent across all 8 directions is the part that always breaks, so this is a big deal. The little wizard walking around that campfire camp reads really well

Thanks for sharing

The first game I tried to make was sprite based and it went terribly, but this looks super clean.

Wow amazing

one character image, eight directions to get distracted by side quests

8 directions from one image is wild 👀

单张图就能补齐八方向,做像素 RPG 省下不少手工活。

one image in, playable character out — that's the part that kills the old pipeline. sprite sheets used to be a week of grunt work, now the bottleneck moves to game design itself. ngl, MCP doing the heavy lifting here is the real flex.

Consistency across 8 directions is the part I would have doubted most. I have watched AI game assets fall apart the moment a character turns, so a pipeline that holds the same character through every angle is a real step forward.

For thr motion control, how it holds up on the agent side, if Claude Code asks for a new attack animation two weeks later is it still the same wizard, or does AutoSprite pin the character somewhere so it doesn't drift between sessions?

Yeah, I'm doing the preliminary work to do a similar experiment, save for the fact that I'm not trying to one shot it from a single image. I'm trying to make a lot more of the individual assets myself while letting the AI handle most of the scripting in Godot. It'll probably end up being a big mess though. At this point this is really more of an educational experience to see which areas I need to focus my learning-efforts.

i'd lose so much time making little characters with this

I tested it this is absolute ….. GARBAGE don’t be fooled or scammed
