I choose a PNG photo. I try to convert to a JPG selecting JPG in File Format. The output stills being PNG. Why?
Another bug: the custom scale in a photo i’m trying to resize is from x0.20 to x24 but I can’t put decimals and x24 doesn’t work either. Maximun for this photo for me is x23 (24840x31050)
Hi there!
I want to download the trial Gigapixel AI on my windows 7 , 64bit.Once downloaded the program doesn’t opens.Can some one tell me what’s going?
thanks in advance
I’m no longer seeing the horizontal banding in gpu mode + max AI model (gtx 1060 with driver 442.59).
If you check here you will see the reason is because W7 is not included in the required specifications
PS you may be aware that AFAIK MS have declared W7 EoL (End of LIfe) so it is no surprise that software developers are not tailoring their software for older Operating Systems.
PPS though having said that both Sharpen AI and DeNoise AI still list W7…will that still be the case as those programs further develop???
Is Gigapixel even using both processing powers at the same time? I was under the impression that it uses either, but not both at once? I also have OpenVINO on all the time, but usually use CPU (GPU mode has some banding issues sometimes)
I don’t believe so because the last time this happened, v4.5.xx I was told that during processing OpenVINO (GPU) will win out.
But I have asked for confirmation of why this is appearing again as previously it was considered an error with calibration in other AI products such as Mask AI. Let’s see what happens with the next release.
Are you aware of GigaPixel resetting the choosen settings after calibration?
Gigapixel 4.6.0 on Windows10
Scaling above 6x, e.g. 6.01x, will produce low quality results. While I could understand that the AI model is tweaked for a maximum of 6x, I can achieve better results by using that and subsequently use other software to upscale further.
I tested it with both the (recommended) CPU mode, and on GPU mode. Same effect.
This has been my first chance to just give v4.6.0 a spin and just out of curious interest I tried my test images without a human face with Face Refinement “on”.
A quick bit of history ~ I reported v4.4.5 crashing under the conditions above and in the end Ashley got me to try Beta v4.5.0.5b and that cured the crashing I had reported.
Well, fast forward to now and tried as outlined above and lo & behold v4.6.0 crashes as did v4.4.5 but not under the Beta v4.5.0.5b
All the discussions and test files are in the discussion thread I had with Ashley…
I look forward to any insights as appropriate.
PS I thought this was discussed with Ashley on the forum but it was on a support Request #205857 for your cross referencing as needed 
Timing tests of Gigapixel on my system indicate that OpenVino is the slowest, then using GPU (significantly faster), and the fastest is having both on at the same time (marginally faster than GPU alone but the improvement is repeatably measurable).
Of course this is specific to my system, and may well be completely different on someone else’s set-up.
Application & Version: Topaz Gigapixel AI Version 4.6.0
Operating System: Windows 10 (10.0)
Graphics Hardware: GeForce RTX 2070 SUPER/PCIe/SSE2
OpenGL Driver: 3.3.0 NVIDIA 442.74
CPU RAM: 16312 MB AMD Ryzen 7 3700X
Video RAM: 8192 MB
Preview Limit: 8000 Pixels
Yes it will.
As far as OpenVINO is concerned the Intel(R) Distribution for Python will run on AMD hardware, I have see benchmarking where, in some cases, there can be inconsistent performance.
I believe also that if you have both selected on AMD hardware it will revert to CPU as was previously explained to me where my i7/GTX 1050/4GB was giving inconsistent results because both GPU & OpenVINO were active.
Once it was corrected, at least for my combination, there was a little difference between both but the OpenVINO option was a little faster.
I cannot imaging that your RTX2070/8GB is slower than the CPU, it just doesn’t seem right as except for the RTX2080 there is nothing that matches its performance. I suspect there may be other issues with GigaPixel processing as, in my case, performance increased on both when the calibration was corrected to select one in v4.5 … the processing using the GPU was 30 seconds (25% faster) and the CPU was 17 seconds faster (10%) for a specific scale factor of 2.5 for a 1600 x 2400 image.
Strange reply AiDon. Maybe it is a language barrier. After calibration both OpenVino and GPU usage are suddenly turned on again. I dunno why. But I can then - as you suggested - choose either OpenVino or the GPU. These does not solve the partially blurry issue, unfortunately.
OK, see this here where i reset to both on and then clicked “Use recommended settings”, then updated the preview to get it to recalibrate:
After recalibrating, leaves it as both:
Reset to GPU on, CPU off and “Use recommended settings”, then updated the preview to get it to recalibrate:
After recalibrating finishes:
So therefore there seems to be a bug, which may not only be in the panel because if I recalibrate and it selects only 1 I get an increase in performance of at least 20-30% on my system which is:
Application & Version: Topaz Gigapixel AI Version 4.6.0
Operating System: Windows 10 (10.0)
Graphics Hardware: GeForce GTX 1050/PCIe/SSE2
OpenGL Driver: 3.3.0 NVIDIA 445.87
CPU RAM: 16269 MB
Video RAM: 4096 MB
Preview Limit: 6423 Pixels
Note that the GPU drivers are the latest also.
Maybe @eric can see why this is so because the benchmarking is only accurate if one option is Off, and also why are we able to select both options even after accurate benchmarking?
On my system I get 20%+ performance increase if it is benchmarked correctly … i.e. selecting one option on and one off before pressing the “Use recommended settings” button to calibrate performance.
I just ran your test of calibration functionality three ways.
GPU On, and OpenVINO Off, Recalibrate and Both are On again.
GPU Off, and OpenVINO On, Recalibrate and Both are On again.
GPU Off, and OpenVINO Off, Recalibrate and Both are On again.
This was always the case with Use Max Quality AI set to either Yes, or No, and Memory Consumption at either Medium or High.
With Memory Consumption set to Low no matter what any of the other settings were before recalibrating it always ended up with GPU Off and OpenVINO On.
I suspect the switching logic in your test was different than mine because you were set to Medium Memory usage with a GPU that has 4Gb of memory, versus mine that has 8Gb.
I checked the various timings again:
Max_Quality_Setting___No_______________Yes_______________Yes
Memory_Setting______High______________High______________Low
GPU_and_CPU_______6.1_sec__________9.16_sec_________13.11_sec
GPU_Only___________5.77_sec_________9.24_sec__________12.9_sec
CPU_Only___________8.2_sec__________15.2_sec__________12.56_sec
Both_Off___________44.37_sec________71.21_sec_________71.04_sec
The fractional second differences were fairly repeatable, but essentially insignificant.
I appreciate that they’ve built in a processing timer to the save function, but I wish it had just a bit more precision than rounded up seconds to show the subtle differences I was measuring with a stopwatch.
Morning AiDon. I repeated what you did to get it calibrated as shown in your post but the final result is still only partially detailed (click to see the full image)
OK, it is time to raise a support request at the main website as it is inconsistent. Go to …
And fill out the form please.
+6.01x is a 2x upscaled.
