AI Asset Pipeline

What It Does ⚡

AI Asset Pipeline is a comprehensive Unity Editor tool that lets you generate 3D models and audio assets using leading AI services — without ever leaving the Unity Editor. Describe what you need in plain text, supply a reference image, or feed in multiple photos, and the pipeline will produce game-ready 3D assets complete with PBR textures, rigging, and animations. An integrated audio generation module lets you create sound effects and music from text prompts using Stability AI's Stable Audio technology.

Whether you're rapid-prototyping, populating a game world, or exploring AI-assisted workflows, AI Asset Pipeline provides a professional, production-grade toolset to accelerate your asset creation.

IMPORTANT — API Keys & Usage Costs ⚠️

This tool does NOT include AI credits or API access. You must sign up with one or more of the supported AI providers and supply your own API keys. All generation costs are billed directly by the respective provider (e.g., Meshy, Tripo3D, Stability AI, OpenAI, Perplexity AI). Typical costs range from 0.50–5 per full-pipeline asset, depending on the provider and post-processing steps you enable.

To help you stay in control of your spending, AI Asset Pipeline includes a dedicated Budget & Wallet window that:

• Queries live wallet/credit balances from every configured provider
• Displays per-provider pricing references so you know costs before you generate
• Lets you set daily budget limits with automatic 80% and 100% threshold warnings
• Tracks per-provider and per-operation usage history over time
• Provides cost estimates directly in the pipeline UI before you click "Generate"

You are always in full control of how much you spend.

Generation Modes 🎨

Text → 3D (Direct) Enter a text description such as "medieval treasure chest" and receive a fully textured 3D model. Supported by both Meshy and Tripo3D.

Image → 3D Supply a reference image — concept art, a photo, or an AI-generated image — and convert it into a 3D model with PBR textures.

Multi-Image → 3D Provide 2–8 images from different angles for photogrammetry-style reconstruction. Ideal when you have multiple reference views of an object.

Full Pipeline (Text → Prompt Optimization → Image → 3D) The flagship mode. Enter a rough text idea. The pipeline:

1. Optimizes your prompt using GPT-4.1 or Perplexity Sonar for maximum generation quality
2. Generates a high-quality reference image via Stability AI, DALL-E, or LetzAI
3. Converts the image into a 3D model with PBR textures

An optional Image Confirmation step pauses the pipeline after image generation so you can preview, accept, regenerate, or cancel before 3D conversion begins — saving credits when the image isn't right.

Audio Generation Generate sound effects, ambient audio, or music from text descriptions using Stability AI's Stable Audio 2.0 / 2.5. Configure audio type, duration (up to 60 seconds), BPM, looping, and output format. Optional prompt optimization ensures best results.

Post-Processing Pipeline 🔧

After initial 3D generation, enable any combination of these optional post-processing steps (each a simple checkbox toggle with inline cost estimates):

• Remesh / Retopology — Cleans and optimizes mesh topology for better edge flow and reduced polygon count
• Retexture — Applies entirely new AI-generated textures using a custom style prompt and optional reference image
• Auto-Rigging — Automatically adds a humanoid skeleton to your model with configurable character height
• Animation — Adds animation clips (idle, walk, run, jump, attack, or any custom animations) to rigged models
• Format Conversion — Converts models between formats: GLTF, USDZ, FBX, OBJ, STL, 3MF with optional quad remesh and face limit control

Most post-processing steps can also be used standalone on existing models — not just on freshly generated ones.

Supported AI Providers 🔌

3D Model Generation

• Meshy — Text→3D, Image→3D, Multi-Image→3D, Remesh, Retexture, Refine, Rigging, Animation, PBR textures
• Tripo3D — Text→3D, Image→3D, Multi-Image→3D, Remesh, Retexture, Rigging, Animation, Stylization, Format Conversion, PBR textures

Image Generation

• Stability AI — Stable Diffusion 3 (Medium, Large, Ultra) for high-quality reference images
• OpenAI DALL-E — DALL-E 3 for creative, artistic reference images
• LetzAI — Flux 2 model with configurable quality, creativity, and resolution settings

Prompt Optimization

• OpenAI — GPT-4.1, GPT-4.1 Mini, o4-mini via the Responses API
• Perplexity AI — Sonar and Sonar Pro models

Audio Generation

• Stability AI Stable Audio — Stable Audio 2.0 and 2.5 for sound effects, ambience, and music

Every provider is a ScriptableObject asset that you configure and assign to a Pipeline Config. The interface-based architecture makes it straightforward to extend with custom providers.

Batch Queue Processing 📦

Generate dozens of assets in a single session:

• Single or Bulk Input — Enter prompts one at a time or paste a list of prompts (one per line)
• Full Pipeline per Item — Each batch item runs through the complete pipeline you've configured (prompt optimization → image → 3D → post-processing)
• Sequential or Parallel Processing — Process items one at a time (safer) or run up to 10 items simultaneously (faster)
• Per-Item Progress — Track status, progress percentage, and current step for every item in the queue
• Queue Management — Retry failed items, clear completed items, cancel processing, reorder the queue
• Automatic Output Organization — Each generated asset is saved to its own subfolder with model files and textures neatly organized

Budget & Cost Management 💰

A dedicated Budget & Wallet window provides full visibility and control over your AI spending:

• Live Wallet Balances — Query real-time credit balances from Meshy, Tripo3D, Stability AI, and others
• Per-Provider Pricing Reference — See exactly what each operation costs before you run it
• Daily Budget Limits — Set a maximum daily spend in USD; the system blocks generation when the limit is reached
• Threshold Warnings — Configurable alerts at 80% and 100% of your budget
• Usage History — Browse detailed per-provider, per-operation usage history with retention up to 30 days
• Token Budget Config — ScriptableObject with full control over daily/hourly/per-provider limits, cost-per-credit rates, rate limiting, and notification settings
• In-Pipeline Cost Estimates — See estimated costs directly in the generation UI before clicking "Generate"

Style System 🎯

Create reusable Style Configurations (ScriptableObjects) to ensure visual consistency:

3D Style Config

• Style keywords and negative keywords appended to every prompt
• Prompt prefix and suffix templates
• Style strength slider
• Color palette references
• Reference images for style transfer
• Preferred polygon density (Very Low → Very High)
• Smooth vs. hard edge shading preference

Audio Style Config

• Style keywords and negative keywords
• Prompt prefix and suffix templates
• Default audio type, duration, and looping behavior
• Tempo (BPM) presets for music
• Genre and mood tags
• Sound effect category tags
• Reference audio clips

Architecture & Extensibility 🏗️

AI Asset Pipeline is built with a clean, modular architecture designed for professional use and easy extension:

• Interface-Based Providers — IModelGenerator, IImageGenerator, IPromptOptimizer, IAudioGenerator, IModelRigger, IModelGeneratorWithCapabilities — implement any interface to create your own providers
• ScriptableObject Configuration — All providers, styles, pipeline configs, and budget settings are ScriptableObject assets, easily shared and version-controlled
• Capability Discovery — The UI dynamically shows and hides features based on the selected provider's declared capabilities (15 capability flags)
• Assembly Definitions — Separate Runtime and Editor assemblies for clean compilation boundaries
• Async/Await Non-Blocking — All API calls use async/await, keeping the Unity Editor responsive during generation
• Event-Driven Progress — Real-time progress reporting with per-step status updates
• Auto-Fallback & Retry — Automatic failover to backup providers and configurable retry counts
• FBX Texture Postprocessor — Automatically assigns PBR textures (albedo, normal, metallic, occlusion, emission) to imported FBX materials when textures are in the same folder
• Encrypted API Key Storage — Keys stored encrypted in EditorPrefs, never in project files

Requirement Details 📋

Unity - 2021.3 LTS or newer

.NET - Standard 2.1

Platform - Windows, macOS, Linux (Editor only)

Internet - Required for API calls

API Keys - At least one 3D provider (Meshy or Tripo3D)

External Dependencies - None — fully self-contained

Documentation and setup guide included 📖