The Complete Local AI Setup Guide
The Complete Local AI Setup GuideEverything you need to run AI image generation, LoRA training, and prompt building on your own PC — no cloud, no subscriptions.Cloud AI tools want you paying monthly whether you use them or not, uploading your images to servers you don't control, and living under content filters that can change overnight. This guide gets you off that treadmill completely.Built from months of hands-on production work — not theory. Every setting, every workaround, every "why is this generating black images" fix in this guide comes from real debugging sessions, real training runs, and real trial and error on consumer hardware.What's InsideCh 1 — Why Local? The case for running everything on your own GPU, and what hardware you actually need (spoiler: you probably already have it).Ch 2 — ComfyUI Setup Installation, folder structure, essential custom nodes, and the launch flags that prevent the most common VRAM headaches.Ch 3 — Models Explained UNet vs DiT architecture, quantization formats (BF16/FP8/NF4/GGUF) decoded, and a full breakdown of Krea 2, FLUX, Ernie Image, Z-Image, and Boogu — what each is good at, what to avoid.Ch 4 — Text Encoders & VAE Why mismatched components produce black or noisy images, and exactly which files pair with which model.Ch 5 — Your First Generation Sampler/scheduler cheat sheets per model, prompt writing for modern DiT models, and resolution rules that actually matter.Ch 6 — LoRA Training on 16GB VRAM Full walkthroughs for OneTrainer, Ostris AI Toolkit, and Kohya-ss — plus a parameter reference chapter (6.1) with copy-paste settings for every tool and model combo. Includes cloud training on RunPod/vast.ai for when local isn't enough.Ch 7 — Multi-LoRA Workflows Which models handle stacked LoRAs well (Krea 2) and which fall apart with more than one (Z-Image) — plus how to avoid identity bleed.Ch 8 — Local Vision Models Ollama vs llama.cpp for prompt generation, and why some "supported" vision models silently fail at actually seeing your images.Ch 9 — Advanced Workflows Instagram carousels with Depth ControlNet, video prompt generation, inpainting across multiple tools, and upscaling pipelines.Ch 10 — Content Pipeline The full production workflow from reference image to published post, with realistic time budgets.Plus two appendices: a troubleshooting guide covering every error you're likely to hit, and a quick-reference cheat sheet for samplers, resolutions, VRAM requirements, and file mappings.Who This Is ForAnyone with an 8-16GB NVIDIA GPU who wants to generate AI images, train character LoRAs, and build a content pipeline — without paying monthly for it or handing their prompts to someone else's server.Format126-page PDF. Instant download.by Nexus
Get it → avgustine4.gumroad.com