I often find I’m not sure how far I can push the sharpening in Photo AI, or whether different models give me a better outcome. Without the 4 windows comparing models as in Sharpen AI, it’s a slow process to evaluate different sets of values. It would be great to be able to save a snapshot of the screen for a particular set of values and just recall the image for comparison instead of having to have Photo AI reprocess the image after a change.
Please select and post under the correct category. Where you posted was in the other than Photo AI category.
And, if you have an idea for improvement, vote for it!
@r-k.green-231014 Thanks for your feedback.
We’re in discussion about an improved preview experience where you can compare different edits and not just the current edit with the original.
1 Like
A feature similar to Photoshop’s snapshot which I can save “snapshots” of image results based on different settings and ability to compare them quickly back and forth or side by side. Currently it can compare between original vs result, but I am taking about extending that to let’s say original vs upscaling by standard vs high fidelity vs low resolution with face detection, etc.
@so.nakagawa Thanks for sharing.
We’ve gotten multiple requests for this and are currently discussing the best way to implement this in a future release!
Gigapixel has a quad setting preview mode so something like that - improved with the option to toggle step thru settings as it is easier to compare at the pixel level when they are in the same place (like eye doctor does not show you A AND B they show you A OR B).
1 Like
The eye doctor analogy is great. I’ll bring it up at our next team discussion.
Note that TPAI already does original or enhanced in place compare, or at least it did as I have not tested the new ui so an extension of that to select compares. IIRC there is or was a pixel shift bug though.
1 Like
The slider that compares model results to the original image is nice. It would be even better to be able to compare the results of two different sharpen models (e.g. Standard vs. Lens Blur) using the slider. It would be even better than the 4-up comparison panels in Sharpen AI, though even that would be a big help in Photo AI. As it is, Photo AI takes time to recalculate every time you change models, which is annoying. Thanks.
I don’t disagree with you at all. The constraint right now is that it would require an extensive overwrite of major sections of the app. I’ll be pushing for it but no guarantees this can get prioritized over the other work that our developers are currently in the middle of.
Using the side by side comparison to see changes done to the original is great, but sometimes I wish I could easily see changes done from one edit to another. (Essentially identical idea to the post “Allow side-by-side comparison with different edits”)
Possible UI solution to this:
Perhaps a button could be added in the lower toolbar to the left of the “Capture” button and the various viewing modes. This button button when pressed creates a "snapshot’ of the current settings, then updates your current viewing mode with the snapshot.
If your’e using the side-by-side or the slider mode, the left preview will be your last created snapshot, and the right would be the currently edited preview where your new changes are applied. The “fullscreen preview” or whatever it’s called would operate the same way, just with the last snapshot being seen when you press or move around the preview.
More buttons:
In addition to the “Snapshot” button at the bottom, above the preview could have some UI as well, like a selector for which snapshot to compare against. A button could be placed in the top left above the preview to select which snapshot (or the original image) to compare to.
As far as I can see, right now the only way to achieve something like this is to create a capture from the button in the bottom toolbar, but this is just too inconvenient a thing to do. Only being able to save it as an image in files, having to open it separately, and not having full quality and maneuverability of the image is not good enough.
Is this to better view different settings rather than one setting compared to the original?
Please add a user preference to enable previous model version use, similar to Video AI allowing it.
Reason: I did some testing and the old high fidelity model (v1.2.10) sometimes gives better results than the current one - it retains more detail, and the file sizes are about 2.5x larger, so that says something about what may be getting lost.
Even better, allow us to select and compare 2 models in the side by side view! This would be a game changer!
Thanks
Correct, to compare edit after edit.
Please consider this feature for upscale and denoise, too. At the moment I help myself by taking screenshots of different models or settings but this is not as elegant as it has been in Gigapixel or the other old apps.
Thanks.
That would indeed be cool.
Another note on that: I noticed that the old 1.3.2 version almost always produces more detailed results, especially in low res situations and low light.
But here it doesn’t seem to be the model, but instead the autopilot settings that newer versions choose (which model for the image and the values).
I did a comparison of three different Photo AI versions with a photo another user had posted in this thread:
If I use the same model and values 1.3.2 detects in 2.x it almost looks the same.
I wish there was some way to let the user tune the current 2.x versions to behave more like 1.3.2 in terms of model selection and value detection. I was hoping that those new weighing sliders would do exactly that, but it’s not the case…
As a result I often find myself processing the same photos both in 1.3.2 and the current version and then doing a side by side comparison to decide, which of the two to keep (I wish it didn’t have to be this way).
2 Likes
I came here to make the same request.
Sometimes it’s difficult to discern whether an incremental change made the image better or worse (especially because the image often takes quite a while to update). Getting an image just right requires a fair bit of tinkering, so the ability to more clearly see the effects of an adjustment would make the whole process much more pleasant and accelerate familiarity with the settings themselves.
I would suggest a toggle that allows the user to choose between “Show Original” and “Show Previous” for the comparison feature.
An extra enhancement would be a “revert to previous” but that would be difficult to implement because it would require holding a complete image version history in memory and performance optimization is already a top priority.
The requested feature to just compare the current output with the immediately previous image would still require some sort of snapshot to be taken as part of the render, but the performance hit would be much smaller and wouldn’t have scaling memory consumption risks like a complete history would.
1 Like
I would like to be able to compare several (2-4) processing variants.
These machining operations should be available quickly (without recalculation) and be comparable side by side.
Thank you very much
It would be really useful to have a way (like old Sharpen) to see the effects of two or three or four sharpening models on the image in a split screen (e.g. quad screen for four chosen sharpen models). I think it’s obvious why this would be useful, just would like to have it as a high priority user-friendly enhancement. Maybe neds to work only on fast GPUs and computers, but it would immensely help selection of the best model for a given image.
That is a great idea! Is there a place to vote for it?
