Setup Guide
- Installing Pinokio on CachyOS
On Arch-based systems, use the AppImage format for the best experience.
Download the .AppImage from the official Pinokio website.
Grant execution permissions: chmod +x Pinokio-X.X.X.AppImage.
Note: If Pinokio fails to launch, ensure fuse2 is installed: sudo pacman -S fuse2.
During the first launch, set the install path to Home. Ensure at least 50GB of disk space is available.
-
ComfyUI Setup and Port Fix
Search for ComfyUI within Pinokio and click Download.
Port Consolidation (Crucial Step): Pinokio often defaults to random ports (e.g., 42000). To sync with RapidRAW:
Launch ComfyUI. In the browser UI, go to Settings (gear icon) -> Server-Config. Set Host to 0.0.0.0, Port to 8188, and CORS header to *. Restart the app in Pinokio (Stop -> Start). -
RapidRAW Integration
Open RapidRAW, navigate to AI Engine settings.
Server address: 127.0.0.1:8188 (Important: Do not include http://).
Fixing “Failed to read workflow file” (OS Error 2): RapidRAW often lacks permissions for /usr/lib/RapidRAW/workflows.
Copy the folder to Home: cp -r /usr/lib/RapidRAW/workflows ~/RapidRAW/. Point RapidRAW to this local folder in the settings menu. -
Hardware Optimization (NVIDIA RTX 16GB VRAM)
To get the most out of your RTX 4060 Ti 16GB and the Flux.1-Schnell model, use these settings in the ComfyUI browser interface (
127.0.0.1:8188):-
KSampler Node (The core processing):
-
Steps: Set to 4. (The Schnell version is optimized for low steps. Going higher to 20+ just wastes time without quality gain).
-
CFG: Set to 1.0. (Crucial for Flux! Anything higher like 8.0 will cause color artifacts and “burnt” images).
-
Sampler/Scheduler:
euler/simple(standard for Flux).
-
-
Empty Latent Image Node:
- Width/Height: Set to 1024 x 1024. (With 16GB VRAM, you have plenty of memory. Flux is a native 1024px model; 512px will look blurry and lack detail).
-
System Level (CachyOS/Linux):
-
VRAM Management: Ensure you are NOT using
--lowvramflags. Your card can fit the entire model in memory, which allows for ultra-fast generation (under 15-20 seconds per image). -
Xformers: If possible, install
xformersvia terminal to further boost speed by ~15%.
-
-