Feedback | Wonder Results With Downscaling

[TPX (Thomas K.)]

Wonder 2 Local.

5,4K → 4X → 21K (319MP)

The cloud doesn’t accept images that large—neither for Photo nor Gigapixel.

The original resolution is a bit too high for the model; it works better if the image is downsized beforehand.

But the result is still very good.
I mean, we’re talking about 21K here.

The background ist HifiV2 - merged via Median filter, the spider was cut out from the Hifi result.

---

Exif: ISO 1600 - f/8 - 1/400 sec. - 1DX MK III - EF 180 mm 3.5L USM - Handheld - 1.5 Meters away.
Software used:
Pure RAW 6 - DxO Deep Prime XD3
Capture One
Photoshop
Topaz Photo (Wonder 2 & HifiV2) + ( Sharpen Portait & Denoise Normal).

---

---

---

---

---

I could try still to throw the spider into redefine.

[Harald.De.Luca]

The details. It’s just enormous. What was the original size of the photo before scaling?

[TPX (Thomas K.)]

Around 5400px x 3600px.

[Harald.De.Luca]

I see. In any case, going from 19.44 megapixels to 319 megapixels with so much final detail is still impressive.

[TPX (Thomas K.)]

Yes, Wonder 2 local is realy nice.

I think i could get more fine detail by downsizing the enlarged picture to its original size and upscale it again.

[Harald.De.Luca]

@TPX

I think i could get more fine detail by downsizing the enlarged picture to its original size and upscale it again.

to test

[TPX (Thomas K.)]

1x Wonder 2 - output Spider size

---

2x Wonder 2 - output Spider size

---

Input Spider size

---

The small image is the original size.

[Harald.De.Luca]

Truly impressive. The Wonder V2 version at 2x magnification really does provide better detail sharpness.

[TPX (Thomas K.)]

Wonder 2 Local + Photo Grain & Sharpening (Wildlife) - Original 24 MP upscaled image.
Not Possible with cloud

---

Wonder 2 Cloud Gigapixel - downsized and enlarged image.

[Harald.De.Luca]

The original is better and the panel in the background is legible compared to the final version.

[TPX (Thomas K.)]

Local processing allows you to preserve details that are not possible in the cloud version.
Because you don’t have to scale down your images.

The downside is its high hardware requirements.
The model uses 16 GB of VRAM and up to 74 GB of RAM.

[Harald.De.Luca]

Yeah, well, enlarging that image size would be impossible on my setup. I only have 8 GB of VRAM and 32 GB of RAM.