When I filed my initial bug report the system posted it to this chat. I thought that was strange however I just continued where the system brought me.
At this point the ball is in your court. You have all the information I am willing to give you. After your QA department follows my steps and still cannot reproduce it please contact me in a way that is less public for you. You have my email, yet you don’t use it? Until then I have done all I am willing to do. I just don’t have time to debug for you anymore.
I’m pretty sure you don’t need my logs to follow the steps. My repo steps make no mention of using logs. Just follow the steps.
As for a public forum, this is the perfect place to have a discussion about this. It helps other users see how Topaz handles issues and their willingness to fix them. At this point I have more dissatisfaction with how my report is being handled than the bug I reported.
Of course, you could just claim it is VRAM again, but, you risk other users running into this bug. I could just decide to stay on v3.3.1 until someday this bug randomly gets fixed or I finally give up and use a different product. It is what it is.
The new Super focus uses Stable Diffusion and is much slower. The other sharpen, denoise, etc. models use Generative Adversarial Networks (GANs). This is the initial part of what chatgpt says:
Generative Adversarial Networks (GANs) and Stable Diffusion are both AI frameworks designed for generative tasks, but they are fundamentally different in architecture, use cases, and underlying mechanisms.
In chatgpt if you just type “gan vs. stable diffusion” you will get a long, very interesting response. I learned something. Here is the last part of the long response:
When to Use Which? Choose GANs for tasks requiring extreme realism, such as face generation or photorealistic textures. Use Stable Diffusion for creative, diverse outputs, especially where textual control or flexibility is needed (e.g., art generation or concept design).
I cannot figure out what the issue is but any time I try to use the face recovery tool, the the image will not process. No preview, and no export. If I remove the face recovery tool, everything processes perfectly fine
Has anyone else had this or a similar problem? I have tried deleting and reinstalling, installing previous releases.
Hi, thank you for all the work going on in the background. But can I please ask for us MAC users when can we expect fixes in both Photo Ai or Gigapixel ?
Photo Ai for example:
And with Gigapixel it pretty much just will not render with any enhancements.
Do you have any approx timescales for fixes please(1 week, 1 month) ?
I’m also on Mac 3,8 GHz 8-Core Intel Core i7, AMD Radeon Pro 5500 XT 8 GB, 40 GB 2667 MHz DDR4 and it’s not possible to make any enlargement, I had to go back to Gigapixel.
The results are all artefacts like a rooftop with tiles looks like just one colour for the whole roof.
As much as we loved our Intel Macs, and as I just mentioned on the Gigapixel thread, they are not up to handling contemporary AI work with any efficiency. Particular results are a different issue, but lack of speed is the main one.
@joseph.vogel - seeing you sent in a ticket to Support, and Josh from the Gigapixel team asked for some info. Was your issue on Photo AI? If so, can you provide the info he requested (files) and mention it was for Photo AI and he will send it our way, and we can have a look at your files to sort this out!
I suspect: Os 12 Monterey Face Recovery Gen 2 incompatibility. This will be resolved in the upcoming v3.4.0 of Photo AI. If that this the OS you are on, open Recover Faces, then switch to Gen 1. You can also update to OS 13 and gen 2 will work. Let me know!
@keith.jones-4783 - Lingyu had a look with the development team and unfortunately, we are not expecting Intel CPU/AMD Radeon GPUs to be able to run the stable diffusion based models such as Remove and Super Focus. These models have higher requirements that the other models, and AMD GPUs on Mac don’t work because of how they are integrated compared to Windows. For example, Windows drivers and Mac drivers are different.
However, all the other models which are not stable diffusion based (Denoise. Raw Denoise, Sharpen, Recover faces, Upscale, Lighting and Color, Preserve text) will all continue to work with your AMD GPU. AMD users on Windows will not have this issue, only Mac AMDs will.
Intel Mac CPUs were always mentioned as underpowered for AI Processing in our System requirements, and this will be even more apparent on stable diffusion models.
@AllMediaLab - can you test if you get artifacts on all AI Processor options in Topaz Photo AI > Preferences > General > AI Processor? Intel Mac CPUs are indeed not ideal and are underpowered for AI Processing. CPU might be the best AI Processor. Let me know.
Hi Mike, CEO Eric reached out to me back in 2007 and I’ve been using and promoting Topaz software since then as a partner, and also volunteer as beta tester. I don’t work for them directly, Dallas is too hot
I see there’s an old paper that discusses some considerations online (at least for video) - wonder what implications for current day sharpening models (if any).
Long range imaging systems that capture video through the atmosphere face a major problem in the form of atmospheric turbulence. This turbulence causes a phenomenon called heat shimmer which appears as a blurring and a wavering geometric distortion of the target scene which limits the effective range of the imaging system.
We explore an image processing approach to mitigating the blurring effect of this distortion by using a blind deconvolution technique to sharpen the video signal and a dynamic illuminance-reflectance correction technique to improve the signal’s contrast. The algorithm is implemented on a Graphics Processing Unit to achieve near real-time performance.
