OBSERVATORY POSTER STUDY

Open the workflow JSON in ComfyUI. This example runs locally. Install CodaComfy Nodes and Comfyroll Studio (CR Apply LoRA Stack) and use a ComfyUI build with MOSS image support.

Use the selectors on the left to choose your local MOSS diffusion model, MOSS text encoder and matching VAE. Re-select the poster LoRA inside LoRA Stack and the compatible eight-step speed adapter inside Load LoRA (Model Only). The saved paths match the demonstration installation. Model weights are separate downloads. When adapting this to another architecture, use that architecture's base model, text encoder, VAE, adapters, sampler settings and latent initializer together.

The published run used MOSS SFT with a cyber-graphic poster LoRA (model 0.65, CLIP 0.60), the HaDeS MOSS eight-step turbo adapter, seed 1732461, 640 x 480, batch 3, Euler/simple, 8 steps and CFG 1. Press Run, click a thumbnail, close the large preview, and choose Continue with selection. The selected image goes to Preview Image. Route selected_latent to a refinement pass when desired.

https://registry.comfy.org/publishers/codalanguez/nodes/coda-comfy-tools
