sdnext-remote

SD.Next extension to send compute tasks to remote inference servers. Aimed to be universal for all providers, feel free to request other providers.

Note

This project is still a Work In Progress, please report issues.

Providers

• (WIP) SD.Next (someone else running SD.Next API)
• (WIP) ComfyUI (someone else running ComfyUI API)
• StableHorde (free)
• (WIP) NovitaAI (paid, affiliated)
• (WIP) ComfyICU (paid, affiliated)
• (WIP) Others :D

Features

	SD.Next API	ComfyUI API	StableHorde	NovitaAI	ComfyICU
Model browsing
Checkpoints browser	✅	🆗	✅	✅	✅
Loras browser	✅	🆗	⭕	✅	✅
Embeddings browser	✅	🆗	⭕	✅	✅
Generation
Txt2img	✅	🆗+	✅	✅	✅
Second pass (hires)	🆗+	🆗	✅	✅	🆗
Second pass (refiner)	🆗	🆗	❌	🆗+	🆗
Img2img	✅	🆗+	✅	✅	🆗+
Inpainting	🆗+	🆗+	✅	✅	🆗+
Outpainting	🆗	🆗	🆗	🆗+	🆗
Upscale & Postprocess	🆗	🆗	✅	🆗	🆗
AnimateDiff	🆗	🆗	❌	❌	🆗
Generation control
Loras and TIs	🆗	🆗	✅	✅	✅
ControlNet	🆗	🆗	⚠️	⚠️	🆗
IpAdapter	🆗	🆗	❌	🆗+	🆗
User
Balance (credits/kudos)	⭕	⭕	✅	✅	✅
Generation cost estimation	⭕	⭕	🆗	🆗	❌

• ✅ functional
• ⚠️ partial functionality
• 🆗+ work in progress
• 🆗 roadmap
• ⭕ not needed
• ❌ not supported

Additional features

• StableHorde worker settings
• Dynamic samplers/upscalers lists
• API calls caching
• Hide NSFW networks option

Why yet another extension ?

There already are plenty of integrations of AI Horde. The point of this extension is to bring all remote providers into the same familiar UI instead of relying on other websites. Eventually I'd also like to add support for other SD.Next extensions like dynamic prompts, deforum, tiled diffusion, adetailer and regional prompter (UI extensions like aspect ratio, image browser, canvas zoom or openpose editor should already be supported).

Installation & usage

1. Installation
   1. Go to extensions > manual install > paste https://github.com/BinaryQuantumSoul/sdnext-remote > install
   2. Go to extensions > manage extensions > apply changes & restart server
   3. Go to system > settings > remote inference > set right api endpoints & keys
2. Usage
   1. Select desired remote inference service in dropdown, refresh model list and select model
   2. Set generations parameters as usual and click generate

Note

You can launch SDNext with --debug to follow api requests

License

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.