ice lab

A Game Development Studio

[huangserva @servasyy_ai]

微软发布了一个 4B模型，可以在 3 秒内把任意图像转换为 3D 资产。

它使用一种名为 O‑Voxel 的新几何格式，并且可以在 CUDA 上用不到 100 毫秒转换为带纹理的网格模型。

输出为 GLB 文件，包含完整的 PBR 材质，可直接用于 Blender、Unity 和 Unreal。

完全开源👇

[Givros @givros]

Welcome if you’re new here 👋

I’m exploring how far you can push AI for game creation: prompts, cinematic scenes, asset sheets, UI concepts, characters, environments and playable browser prototypes.

Most of what I share is built with Codex, GPT Image 2.0 and GPT-5.5.

You can support the experiments here ☕
https://t.co/gxg1bCY439

[Tom Dörr @tom_doerr]

Three.js particle system with visual editor

https://t.co/BJacpBPT2r

[Thomas Ricouard @Dimillian]

I've asked Codex for simple polygons but it seems like it cooked way beyond that goal

[Tom Dörr @tom_doerr]

Allows AI agents to control Unity Editor

https://t.co/mdLxXbGxQy

[BubbleBrain @BubbleBrain]

This is cool. by gpt image 2

---prompt---
A collection of scattered printed photo snapshots placed on a wooden table, showing a series of candid moments of a fictional adult East Asian woman alone in a hotel room at night. Photorealistic, Japanese negative film aesthetic, muted tones, low contrast, subtle greenish tint, soft highlight bloom, visible film grain, direct on-camera flash. Subject: adult East Asian woman, short black hair, natural beauty, calm and slightly distant expression with subtle sensual allure, no resemblance to any real person. Outfit: seductively unbuttoned white shirt revealing a hint of cleavage and delicate lace bra, black mini skirt riding slightly up her thighs, relaxed styling with natural wrinkles in fabric, subtly disheveled in a tasteful yet alluring way. Scene: small hotel room, bed with patterned sheets, bedside lamp, curtains partially closed, night city lights faintly visible through window. Moments captured: * lying on bed in a relaxed seductive pose, body slightly arched with one knee bent invitingly * sitting on bed looking down, legs elegantly crossed to accentuate her curves, subtle arch in her back * leaning back against the headboard with arms raised above her head, back sensually arched, shirt pulling open at the chest * eating simple food at a table, slowly bringing a bite to her lips with a sensual touch * drinking from a glass, head tilted back slightly, lips sensually pressed against the rim * turning back over her shoulder with shirt slipping off both shoulders, looking back with a mysterious distant gaze Composition: each image is a candid snapshot, imperfect framing, slightly off-angle, natural and unposed feeling. Lighting: harsh direct flash combined with dim ambient tungsten light, creating strong contrast and a nostalgic disposable camera look. Arrangement: multiple printed photos randomly scattered on a wooden surface, overlapping slightly, creating a narrative collage. Mood: quiet, intimate, slightly lonely, late-night solitude, everyday realism, subtle sensuality without explicit exposure. 4K, ultra-detailed, photorealistic, no text, no watermark.

[Shaders Bible Series @ushadersbible]

Just 428 more people and we start the project! Anyone interested in real-time water simulation in Unity? Vote here ➡️ https://t.co/RgfSpn6dJY

#gamedev

[Ciri @Ciri_ai]

GPT Image 2 on Chatgpt

Prompt: LUltra-realistic iPhone fast-food restaurant photo, identity preserved exactly from reference image.

Young woman sitting inside a McDonald’s-style fast food restaurant booth, photographed from across the table with a slightly wide iPhone angle. She is centered in the frame, sitting behind a messy spread of fast food on the table.

Pose:
She holds a brown McDonald’s paper bag on top of her head with both hands, elbows raised. She is making a playful funny expression: one eye winked, lips pushed into a pout, slightly annoyed but cute. Her body leans back against the booth.

Outfit:
White fitted sleeveless top, long dark loose hair, delicate necklace or chain, long nails, casual cute styling.

Table details:
The foreground table is filled with McDonald’s-style food packaging: fries box, burger wrapper, paper cups with straws, chicken box, sauce packets, tray liner with colorful anime-style graphics, brown takeout bag, messy fast-food spread. The food and packaging should feel real and slightly chaotic.

Environment:
Modern McDonald’s-style interior with tiled white wall, dark grey brick wall, black wall lettering, booth seating, small tables and stools in background, overhead fast-food lighting.

Camera + lighting:
Raw iPhone indoor photo, slightly harsh overhead light, realistic shadows, casual social-media snapshot, not polished.

Important:
Match the seated booth framing, bag held on top of head, wink/pout expression, white sleeveless top, messy McDonald’s food spread in foreground, restaurant wall graphics, and funny candid energy closely.

Natural skin texture, realistic food packaging, no beauty filter, no smoothing. 4K vertical 9:16.

[BIBI一姐（重开版) @bibisister0508]

照片好看，视频喜欢吗💕😘

[TechHalla @techhalla]

Ever seen AI 2D animation this clean? Not guessing poses anymore!

I used proper movement sheets as reference on Magnific and it changed everything. Here's how 👇

[TachikomaRed @smolekoma]

This will save hours of work for pixel artists 🦀

Built a free, 100% client-side pixel art fixer in Rust → WASM

Drop any AI-generated pixel art and it:
• snaps every pixel to a perfect grid
• removes backgrounds (flood fill or global color)
• converts sprite sheets → animated GIFs
• bulk processes a whole folder

support building $TACHI

[Chong-U @chongdashu]

@langwish_app I'm using a modified version of https://t.co/9EixcQXTTD

[Chong-U @chongdashu]

If you use AI to generate pixel art

You'll know that while the frames look good...

They are often times not aligned -- making your character slide all around.

As part of my suite of tools to fix this, I vibe coded a tool to help me align frames to ensure this doesn't happen.

[Stijn Feijen @spwfeijen]

Claude + MkaeUGC = 550 slideshow posts per day

Little pro tip:

Have your models an iPhone to cover the face.

This makes the AI look even more real

[Sakura Rabbit 樱花兔 @Sakura_Rabbiter]

Mesh blend~🥳

#unity #shader #realtime #gamedev

[亀井りく｜「RecGame」広報担当 @RecGame_RA]

UE5でメッシュとメッシュの境界線を 自然に馴染ませるツール「MeshBlend」、
これは地味にすごいやつです…👀

岩が地面から生えてるように見える質感、 ずっと眺めていられますね。インディーからAAAタイトルまで 実際に使われているのも納得です。

[0xFunky @0x0funky]

1小時多，用嘴巴開發做了一個類 Kingdom Rush 的塔防小遊戲。

全程使用 Codex + agent-sprite-forge skill，從地圖、Props、塔、怪物動畫、Boss、飛行怪、HUD，到 Godot 裡的波次、難度、升級 / 賣塔流程，基本上都是跟 Codex 對話做出來的。

比較有感的是，這次不只是生素材，也不是單純用 JS 做一個陽春版 Demo，而是一路把素材接進 Godot 專案裡，真的用遊戲引擎把地圖、角色、塔、防守波次和互動流程串起來。

當然 Codex 目前還沒辦法非常精準地處理所有位置資訊，像是地圖偏移、物件對齊、塔的位置這些細節，最後還是需要在 Godot 裡手動微調。

但整個手動調整時間不到 10 分鐘。

Codex + agent-sprite-forge + Godot 這條 AI 2D 遊戲開發路線，基本上已經打通了。

接下來準備再來試試看 Unity。

[this could be us @loveucouple]

need

[姫乃かなで @himekanade_]

太ももどあっぷ♡