# ComfyUI / Invoke

Images you generate. ComfyUI is GPL; Invoke is Apache-2.0 — Midjourney without the render in someone else's Discord.

**Replaces:** Midjourney

**Tier:** Medium

## Problem

Midjourney is a hosted image model behind a Discord or web seat. You wanted a graph or a studio that talks to weights you run. ComfyUI is GPL (node graph). Invoke (Apache-2.0 Community) is the studio UI. You still bring models and, usually, a GPU. This is not Midjourney's every style or community feed. Railway GPUs are optional and often the wrong box — a workstation or a GPU host is the honest path.

## Stack

- [ComfyUI](https://github.com/comfyanonymous/ComfyUI) — GPL node-graph image generation
- [Invoke](https://invoke-ai.github.io/InvokeAI/) — Apache-2.0 Community studio UI if you want that editor

## Estimated SaaS cost saved

| You stop paying | Typical / year | Self-host instead |
|---|---:|---|
| Midjourney paid (small team) | $1,200 | ComfyUI or Invoke + weights |

**About $1,200 / year** on a small team, before you count lock-in.

## Deploy on Railway

Placeholder: ComfyUI or Invoke plus a volume for models. A GPU helps; CPU is slow. No one-click template in this repo yet — Start on Railway. Chat UIs stay on /stacks/open-webui-litellm.

[Start on Railway](https://c17.io/go/railway/comfyui-image?from=%2Fstacks%2Fcomfyui-image.md)

[Open WebUI if this is chat, not image generation](https://c17.io/stacks/open-webui-litellm)

## AGENT_PROMPT

```
Replace Midjourney with ComfyUI (preferred) or Invoke Community.

- Persist models on a volume. Bind the UI. Generate one test image. Do not claim Midjourney style or Discord-bot parity.
- GPU is recommended. Write AGENT_PROMPT.md.
```

## FAQ

### Will this match Midjourney?

No promise. You run open weights. Eval the models you actually load.

### Why not Open WebUI?

Open WebUI is chat and an API proxy (/replace/chatgpt). This page is image generation.

Source: https://c17.io/stacks/comfyui-image