Models, Engines & Self-Hosting

Running it yourself — our CPU pipeline, your PC, Colab, Modal, bringing your own key. RAM/VRAM, times, quantisation, and what CPU-only genuinely cannot do.

Belongs here: The command you ran and what it cost you in minutes.

Goes elsewhere: Model downloads and LoRA hosting belong on CivitAI and Hugging Face; link out and discuss here.

Related: Your engines

＋ New thread

How much VRAM do you need? SD 1.5, SDXL and Flux, honestly

by @hathor · 2026-10-19 10:12 UTC

✦ 0 · 1 replies
2026-10-19 13:12 UTC

What CPU-only image generation genuinely cannot do (and what it can)

by @hathor · 2026-10-19 12:12 UTC

✦ 0 · 0 replies
2026-10-19 12:12 UTC

Driving ComfyUI from a script: the HTTP API in five calls

by @hathor · 2026-10-19 11:54 UTC

✦ 0 · 0 replies
2026-10-19 11:54 UTC

Quantisation for image models: fp16, fp8 and GGUF, in plain terms

by @hathor · 2026-10-19 11:36 UTC

✦ 0 · 0 replies
2026-10-19 11:36 UTC

Bring your own key: fal.ai or Gemini in the Studio, and what happens to the key

by @hathor · 2026-10-19 11:18 UTC

✦ 0 · 0 replies
2026-10-19 11:18 UTC

Running the Studio's engine on your own machine or a free Colab

by @hathor · 2026-10-19 11:00 UTC

✦ 0 · 0 replies
2026-10-19 11:00 UTC

ComfyUI from zero: the seven nodes of a basic workflow

by @hathor · 2026-10-19 10:36 UTC

✦ 0 · 0 replies
2026-10-19 10:36 UTC

LCM on CPU: why Hathor Studio takes one to three minutes, and the settings behind it

by @hathor · 2026-10-19 09:48 UTC

✦ 0 · 0 replies
2026-10-19 09:48 UTC