This is a quick update to address some issues related to the last release of Gigapixel when loading RAW or monochrome images.
You may also notice the hardware support for Qualcomm Snapdragon CPUs, NPUs, and NVIDIA RTX 50-series cards below. If this applies to you - make sure to download the appropriate build.
You can use the in-app updater or download using the links below:
v8.3.2 Released March 24, 2025 Mac:Download Windows (Includes RTX 50-series support):Download
Here’s a special build for the new Qualcomm CPUs out there:
@william.lobb sorry to hear that! Can you share the JPG image causing the error with the support team or post it here if you haven’t already. Collecting any remaining examples. If you have any logs you can share that may help as well.
Upgraded from 8.3.1 to 8.3.2 and it still crashes without an error message displayed, when dragging and dropping a jpg file to it. Tried several different jpgs with the same results.
My installation crashes regardless of the file type or the source of the file. I’ve tried jpgs from a Nikon camera and a Pixel phone, and Nikon RAW files (.NEF). They all crash…
Hey Susan! Can you please open a support ticket at support@topazlabs.com and send us your application logs there? This will help us prioritize the fix. Looking into it.
I have 6.0.0, but I want to download the trial of the current version to see if I want to upgrade. Will it overwrite 6.0.0? If I decide to not upgrade will 6.0.0 still be there and usable?
Still crashes without error messages when I try to open any kind of images. No matter if is a jpeg, tiff, or something else. It just closes itself without a notification. 13th Gen Intel(R) Core™ i7-13700KF 3.40 GHz with Nvidia RTX4070 latest drivers.
<Error>
<Code>RequestHeaderSectionTooLarge</Code>
<Message>Your request header section exceeds the maximum allowed size.</Message>
<MaxSizeAllowed>8192</MaxSizeAllowed>
<RequestId>VXK9HYQMM6REFT0A</RequestId>
<HostId>RCKMn5OkhFg5g/F4y57pGmy4zWVVfwL9USyYvWqs3SQIysZujYft9ww4dpTrGI+IDrFco+/LFV3jZ4Ho/LKfvWaPj+l1xt55</HostId>
</Error>
Hi, it looks like the header for the download is too large.
Following the v8.3.2 Windows (non-ARM) download link without query params seems to work for me. …/deploy/TopazGigapixelAI/8.3.2/TopazGigapixelAI-8.3.2.msi
Even after the update (v 8.3.2) Gigapixel crashes every time I try to load a new file - completely independent of the source data format (jpg/png/tiff) or the color space. It also doesn’t matter whether you use the “open image …” dialog or drag and drop images into the workspace.
Hi, I purchased the latest version of Gigapixel less than half an hour ago. I’ve installed it, but none of the models appear. I’d like to know why this is happening and, above all, how to fix it. Since I paid for it, I want to use it. Here’s a photo and log files.
Would it be feasible to provide Redefine with a setting that adds less sharpness to the output image than the current version?
I do mainly photorealistic AI generated animal pictures that usually have light artefacts to them such as combing/regular patterns in the fur textures. Redefine seems to have great potential in doing away with those artefact, giving the image a much more realistic appearance. That is, if the Redefine result were not as over-sharpened as it is now.
I can get excellent results on people using creativity 2, texture 1 but with animals, the results are just too sharp. Tiny hairs come out as needles. Cranking creativity down to 1 doesn’t make things better because at 1, prompting seems not possible, seemingly making the AI “guess” what is in the pic, thus messing up the fur texture even more than creativity 2 with an adequate prompt.
I have attached the Redefine result (top picture) and the original out of Stable Diffusion (bottom picture) to this post. As you can see, the potential of Redefine seems huge but the sharpening is just too much. Running the result through High Definition afterwards with a high upscale factor and denoise cranked up to 100 makes it somewhat smoother but it would be much more convenient to have a less over-sharpened output directly from Redefine.
