2-step pDMD LoRA for p-h3-image released, fix blurry fast-generation outputs
Illustrious_Row_9971 · reddit · 2026-11-07
A Reddit user released a 2-step pDMD LoRA for p-h3-image, hosted on Hugging Face (akhaliq/p-h3-image-pdmd-2step).
Side-by-side comparisons show the base model produces blurry, degraded images at 2 steps, while adding the LoRA yields sharp, detailed results across ukiyo-e, anime, photorealistic portraits and sticker styles — a dramatic quality jump at the same step count.
More from Multimodal
- Character design sheet generated with Adobe Firefly — azed_ai · 2026-11-07
- SCAIL-2 vs MiniMax for motion control: 5s 720p takes ~25 min on RTX 5090 — DragonfruitNo9822 · 2026-11-07
- Dev Adds Force Effectors Like Wind to Gaussian Splatting in Blender — 5agado · 2026-11-07
- Neon fox video generated with Seedance 2.5 showcases Higgsfield's video model — CurieuxExplorer · 2026-11-07
- Open-source OmniVoice Studio dubs video in 646 languages fully local, taking on ElevenLabs — thisdudelikesAI · 2026-11-07
- Redditor turns mom's dream-house description into a touching AI render — FirstOil3672 · 2026-11-07