YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

✨ PGUpscaler

Opraví plastovú AI pleť a fotku zväčší a doostrí. · Fixes plastic AI skin, then upscales and sharpens the photo.


🇸🇰 Slovensky

  1. Pods → + Deploy, GPU 24 GB+ VRAM (RTX 4090 / 5090; na zväčšenie 3× radšej 48 GB), filter CUDA Version 13.0+.
  2. Change Template → PGUpscaler → Deploy On-Demand, počkaj na Running.
  3. Connect → HTTP Service → Port 8188. Ukáže sa stránka s priebehom. Prvý štart trvá 5 – 10 min (~23 GB), potom sa sama prepne do ComfyUI.
  4. Workflows → PGUpscaler - SK.
  5. Do 1️⃣ SEM NAHRAJ FOTKU nahraj fotku a klikni Run. Prvýkrát nič iné nemeň.

Nastavenia (🟣 TU MENÍŠ):

  • 2️⃣ SILA – 0.2 jemne · 0.3 default · 0.4 silno
  • 3️⃣ IBA TVÁR? – false = celá fotka · true = iba tvár
  • 4️⃣ ZVÄČŠENIE – 1.5 · 2 default · 3 (veľmi pomalé)

Výsledok: ✅ HOTOVÁ FOTKA sa uloží do output/plet (pravý klik → Save Image). Viac fotiek naraz: Port 8080 (FileBrowser, admin / adminadmin12).

Keď niečo nejde:

  • Pod sa nespustí → filter CUDA 13.0+.
  • „502“ → počkaj 1 – 2 min, F5. „Niečo sa pokazilo“ / „CHYBA“ v Logs → ⋮ → Restart Pod.
  • Iná tvár → SILA 0.2. Plastová pleť → SILA 0.35. Out of memory → ZVÄČŠENIE 1.5.
  • Run nič nerobí → zmeň fotku alebo nastavenie. Vlastné zmeny ukladaj cez Save As.

Po práci daj Stop (platí sa len disk), Terminate zmaže aj disk.


🇬🇧 English

  1. Pods → + Deploy, GPU 24 GB+ VRAM (RTX 4090 / 5090; 48 GB for 3× upscale), filter CUDA Version 13.0+.
  2. Change Template → PGUpscaler → Deploy On-Demand, wait for Running.
  3. Connect → HTTP Service → Port 8188. A progress page opens. The first start takes 5 – 10 min (~23 GB), then it switches to ComfyUI by itself.
  4. Workflows → PGUpscaler - EN.
  5. Upload your photo in 1️⃣ UPLOAD YOUR PHOTO HERE and click Run. First time, change nothing else.

Settings (🟣 CHANGE HERE):

  • 2️⃣ STRENGTH – 0.2 subtle · 0.3 default · 0.4 strong
  • 3️⃣ FACE ONLY? – false = whole photo · true = face only
  • 4️⃣ UPSCALE – 1.5 · 2 default · 3 (very slow)

Result: ✅ FINAL PHOTO is saved to output/skin_fix (right-click → Save Image). Many photos at once: Port 8080 (FileBrowser, admin / adminadmin12).

When something goes wrong:

  • Pod doesn't start → filter CUDA 13.0+.
  • “502” → wait 1 – 2 min, F5. “Niečo sa pokazilo” / “CHYBA” in Logs → ⋮ → Restart Pod.
  • Different face → STRENGTH 0.2. Plastic skin → STRENGTH 0.35. Out of memory → UPSCALE 1.5.
  • Run does nothing → change the photo or a setting. Save your own changes with Save As.

When done click Stop (only the disk is billed), Terminate deletes the disk too.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support