BennyDaBall/PiD-1.5-qwenimage-nvfp4-comfy
🚀 PiD 1.5 qwenimage — NVFP4 (ComfyUI-ready)
The first quantized PiD decoder on HF. 2.61 GB → 1.09 GB, drops straight into a stock UNETLoader, and decodes a 33 MP master in 16 s on a 5090 (bf16 takes 20). No custom loader. No core patches. No Blackwell requirement either — this runs anywhere the bf16 file runs.
🧠 What is this?
NVIDIA's PiD v1.5 decoder for the qwen-image latent family — the 4-step pixel-diffusion decode that replaces your VAE decode and hands back 4× the resolution. Same decoder, 42% of the size, a little faster.
Everyone else ships this model in bf16 only. Not anymore.
👀 Same seed, same latent, both builds
100% crops — faces, lettering, bokeh all hold:
📋 Specs
💻 Usage
Drop the model in ComfyUI/models/diffusion_models/, select it in UNETLoader, done. Two workflows included:
Simple (core nodes only) — PiD-1.5-qwenimage-nvfp4_simple_1mp_to_16mp.json. Stage-1 render at ~1MP → single-shot PiD 4× → 5376×3072, built from nothing but core nodes (PiDConditioning + a 4-step KSampler). Drop the included pixel_space_vae.safetensors into ComfyUI/models/vae/ for the final decode step. Keep this one near its trained envelope — single-shot PiD collapses past ~4K output.
Tiled (the big guns) — PiD-1.5-qwenimage-nvfp4_tiled_2mp_to_33mp.json. Stage-1 → seam-free 7680×4352 in one queue via ComfyUI-Latent-Tiled-PiD: install from ComfyUI Manager (search "Latent-Tiled-PiD") or the Comfy Registry.
Every included workflow was executed through the actual ComfyUI frontend against this exact file before upload. qwen-family latents only — the FLUX.2 build is here.
⚖️ License & credits
NVIDIA created PiD (paper); Comfy-Org published the bf16 ComfyUI repackage this converts. NSCLv1 — non-commercial research/evaluation use only, derivatives included. Read the nvidia/PiD card before you build anything on it. Not affiliated with NVIDIA or Comfy-Org. Conversion and validation by BennyDaBall_OG.