Topaz Photo AI v1.0.5

[Martyprod (Marty)]

hello, sorry it has certainily said a billion of time just wanted to tell

• would love a zoom by default, must all the time select “zoom to fit” and when i save the picture, the picture come back to 100%
• does a crop feature / fonction is planned at some point ?
• lot of vertical bars on a picture i’m trying, but it’s a known Ai issue if i remember.
• agree, hope at some point a possibility to manually select masking because sometimes, it fail to find the right “subjects”.
• I miss a “+” / add button, somewhere, once we added a picture, we must right click /add , or file / open.
• the height by default is not working i put a height by default on 2160px, it process the picture (640x384) and then, it put the picture at full screen 271% … impossible to use it’s related to the enhance resolution. it does a 2160 upscal with enhance turned off, and then it turn it on making the picture to be 271% big with only a 100% minimal zoom…

Thanks !

[alexatkin]

Unless its different in this application, these are the scaling factors there are AI models for so recommended for best results. Anything else would involve none-AI resizing after the AI pass, which is best done manually to ensure you aren’t degrading the result.

It actually does strangely do arbitrary sizes by default with autopilot. Can a software developer weigh in on the implications of autopilot working this way now? I can only see that being worth it if it always uses the next model upwards and scales down, which Gigapixel did not do so has this been changed in PAI?

[david.123 (New Profile Picture!)]

Makes sense. But still no reason to force AP selected resize filters on the users. We should have complete control over whether natural/graphics/etc is applied to resized pictures. And same for noise reduction.

[Fotomaker]

@Ulrich_Pohl

Hi, what happens when you open more then one Canon CR2 File in Stand Alone Modus? Do you see thumbnails in the lower left corner?
I don´t (Win10) and reported this issue a few time.

Hi Ulrich,

At your request, I just opened a series of Canon CR2 raw images into my PAI 1.0.5 Standalone (Win 11) to see what would appear. I don’t typically process batch. So tried it to see what would happen.

I do Not see any pictorial image previews in the thumbnails at the bottom of the PAI workspace where the icons for the the batch of opened CR2 (Canon raw) images appear. I only see icons with symbols in them. That would be useless to me if I had to be able to click on specific images to make adjustments. It also does Not tell me if the correct set of images opened b/c I don’t see small thumbnails of the actual images - just a rectangle with symbol. Here’s a snip of what I see:

Also, as @PaulM experienced (I assume still with a Win 10, though I’m now on Win 11) there is a complete Mismatch b/wn what appears as the settings for the batch image icons at bottom of the workspace (when hovering over them) & what appears in the righthand PAI Settings Panel under Sharpen.

See 2 snips below.

Sharpen appears to be completely messed up! It:
1- Selects (using A.P. & in the righthand panel) Standard for every single image of any format (raw, jpg, tiff) - per righthand Settings Panel. (I only tried batch at the rqst of another user - b/c I only process single images, typically, all I ever see is “Std” model applied in this rel.)
2- Uses only Motion - Level 9 (in the notation for each opened batch image) for every single opened image - even though there were test shots being processed by A.P. that contained motion by animals from different times by diff. animals moving in diff. ways and diff. spds. It seems awfully interesting that they’d all require the exact Level 9 Motion correction…
3- Shows notations for Sharpen (Denoise matches) that are diff. at bottom of the UI & in the righthand Settings panel…
4- Doesn’t sharpen the image in the workspace in an acceptable manner. (Clearly, I know to wait until the processing is completed…).

[alexatkin]

Its not forced, you can change it by specifying an exact resolution just as you can change the AI model used by opening up the options for the specific function, so not sure what you feel is missing here?

By definition, Autopilot is intended to try to auto-detect the best settings.

My biggest bugbear right now is not being able to right click on the preview to close the image/all images.

[david.123 (New Profile Picture!)]

@alexatkin

Its not forced, you can change it by specifying an exact resolution just as you can change the AI model used by opening up the options for the specific function, so not sure what you feel is missing here?

I’m talking about the ‘enhance resolution’ filter that we have no control over and almost always defaults to the ‘low resolution’ setting. There’s no good reason for us not having control over this in the AP configuration.

[Martyprod (Marty)]

• for an unknow reason, masking refine stop working sometimes… the picture stay at “view original” and nothing i can do about it… moving slider has no consequencies on the mask.
• having the default of enhance on low resolution is … like @david.123 told, at least having the possibility to put by default the one we want but does the detection of the picture take in consideration the upscaling ? it’s like the recover faces which is all the time at 100. most of the time, i must decrease it to around 30 to 40%

[AiDon]

You need to check on the LibRAW website to see if they are supported, or contact support.

[alexatkin]

Its not a “default”, its Auto-detection. It might be flawed but ideally we want it improving rather than forcing an override.

How are we ever going to know if autopilot has improved if we start just overriding its settings by default? The size override makes sense for the reason I mentioned, fixing it to the optimum model size. I find thatg using over 2x is never better whereas changing the model based on the input definitely does make a huge difference.

That said, I do agree low resolution is not the best option in a lot of cases, but forcing it to something else you’d still have the same problem in that on many images it would be using a sub-optimal model unless you manually change it per image.

This is why I prefer Gigapixel still, PhotoAI is really not retail ready in my opinion.

[david.123 (New Profile Picture!)]

@alexatkin

How are we ever going to know if autopilot has improved if we start just overriding its settings by default? The size override makes sense for the reason I mentioned, fixing it to the optimum model size. I find thatg using over 2x is never better whereas changing the model based on the input definitely does make a huge difference.

I’ve already suggested that AP configuration options be provided for selecting which enhance resolution filter is applied to upscaled images including an ‘auto’ function if one wants to keep it auto detected. ‘Natural’ works best for my purposes so I should be able to set that as a default. It would be very simple to include that functionality in the AP configuration.

You can feel free to use AP for everything. I however prefer control over my software.

Understand?

[31Fairlane]

@adam.mains

Lower memory footprint of TIFF/DNG output (6000x4000 saves ~48MB, 32000x32000 saves ~2GB)

I Don’t see improvement with output file sizes.
Saving an ORF RAW file (20MB) results in a DNG of 121MB or a TIFF of 103MB
For reference … DXO PhotoRaw outputs DNG of around 60-75MB…still large but about half of the Topaz products.
PR 1.0.5

[Fotomaker]

PAI 1.0.5 Standalone. Win 11 Pro. Processor = Auto (all specs of mine above in earlier posts)

I tried/posted rpt of experiences with CR2 raw images (single & batch) in A.P. in the past 2-3 days.

Today I tried a Sony ARW raw batch to see if it was the brand of raw that was the issue. It was not a branded raw images issue.

The issues are with PAI coding and are consistent across raw images I ‘try’ to process using PAI 1.0.5 Standalone. I reported the problems that developed in the 1.0.5 rel. Ps Plugin separately above…

Issues that are Repeated Using ARW Raw images on A.P.:
1- No visual images appear on the icons for the imported ARW image batch at bottom of the UI. I only see icons with symbols. That is not useful. Must see the images to be able to go back to specific images to tweak settings & to verify correct images were uploaded.

2- I get a Mismatch of settings applied by A.P. (for Sharpening) when hovering over any given batch ARW image icon at base of the UI & what is shown as selected by A.P. in the righthand Settings Panel. I’ll see a Motion model # used at bottom of the UI & it’s always the Std Model that’s active (incl. with the green dot indicator) in the righthand Settings Panel & a different value is used for Std.

3- My image that’s in the PAI workspace remains typically soft-looking. Doesn’t look like a sentient robot being (A.P.) has made it sharper (Contrasty?) or clearer (Clarity).

4- When I manually change the Sharpen setting (in the righthand settings panel) to match the setting the hover at bottom of the UI shows then I get an acceptably sharpened image (ie, if my hover shows Sharpen - Motion = 49 & the righthand panel (always!) shows Standard & I manually change the righthand panel to be on Sharpen - Motion Blur Model = 49 then the image looks processed as it should be).

Whatever is triggering Standard Model to be the std A.P. setting in all instances (including Single & Batch image processing) is what’s preventing the images from being processed/sharpened correctly. I don’t program. I can only use logic. It seems like some “If, then…” parameter is screwed up. So far, , I am not seeing the same type of issue with Denoise - it clears the noise and the settings are in sync from batch thumbnails (bot. of UI) to righthand panel.

Latest Snips from Today’s Test Using Sony ARW Raw Images (in batch):

A.P.s Current Mismatch Settings Effect - Sharpen Motion Blur Model (base of UI on Batch Thumbnails) & Sharpen Std Model in Settings Panel —> Produces Blurry Processed Preview

Manual Override to Make Settings Match As They Should (i.e., Motion Blur shown in ARW thumbnails manually made to show in righthand panel to override the “Std” A.P. generated) —> Image Preview Looks Clear and Acceptably Sharpened:

[e_dinur]

@31Fairlane

@adam.mains

Lower memory footprint of TIFF/DNG output (6000x4000 saves ~48MB, 32000x32000 saves ~2GB)

I Don’t see improvement with output file sizes.
Saving an ORF RAW file (20MB) results in a DNG of 121MB or a TIFF of 103MB
For reference … DXO PhotoRaw outputs DNG of around 60-75MB…still large but about half of the Topaz products.
PR 1.0.5

If you want to know the full (uncompressed) digital volume of a sixteen bit RGB image flle in MB, multiply the MP by 5.7. Thus, in the example above 4000x6000 = 24 MP x 5.7 = 137 MB. Compressing that down to 48 MB sounds pretty substantial to me and even makes me wonder what the cost is.
BTW, for a 14 bit Raw file the magic number is 1.7, so a full 24 MP Raw is 40 MP. Anything less than that has been compressed.

[matt-9806]

Would be very nice IF one could tell Photo AI where there is a face when the AI does not see it - because of something on the face like a finger in the mouth, or a partial cover of the face. The option to chose it and then Photo AI do its magic.

[Fotomaker]

Consistently getting this error msg for one of my ARW images … never seen this before:

[AiDon]

Typically, that message appears if you are trying to open an empty image. Check with another app to see if it can be opened … don’t use Windows Photo viewer as that will show the thumbnail.

[plugsnpixels]

Crash on quit, M1 MBP, latest Ventura beta:

Crashed Thread: 30 Thread (pooled)

Exception Type: EXC_BAD_ACCESS (SIGSEGV)
Exception Codes: KERN_INVALID_ADDRESS at 0x000000016e1626a0
Exception Codes: 0x0000000000000001, 0x000000016e1626a0

Termination Reason: Namespace SIGNAL, Code 11 Segmentation fault: 11
Terminating Process: exc handler [61970]

VM Region Info: 0x16e1626a0 is not in any region. Bytes after previous region: 4548257 Bytes before following region: 39264
REGION TYPE START - END [ VSIZE] PRT/MAX SHRMOD REGION DETAIL
Stack 16dc84000-16dd0c000 [ 544K] rw-/rwx SM=PRV thread 7
—> GAP OF 0x460000 BYTES
Stack Guard 16e16c000-16e170000 [ 16K] —/rwx SM=NUL

[Fotomaker]

I see the image in my Win Explorer, but I’ll try something else too! It’s from a set of images others of which work.

[Fotomaker]

@AiDon

Typically, that message appears if you are trying to open an empty image. Check with another app to see if it can be opened

I just tried the Sony ARW that caused the error msg in PAI 1.0.5 Standalone with Ps ACR - it launched okay in that & then opened okay into Ps.

Once it was in Ps I was able to open it in PAI Plugin okay (I realize it’s not seeing it as a raw anymore at that point, but still …). It was able to process with A.P. & by me making manual changes to sharpen the face & details more than A.P. did. Interestingly, PAI detected not only the guy’s face, but also the face of the Black antique doll in front of him (with the red hat). It saved back to the Ps launch layer okay too:

Very odd. I consistently got that error msg using the PAI standalone with that Sony ARW image & also another one of the same guy, just a diff. shot. Maybe that one doll is a voodoo doll and it jinxed the app!!!

[plugsnpixels]

Why I love enlarging my old pix with Topaz apps! From a scan of a 5x7 print of a 1980s 35mm negative.

The diagonal lines are rain effects (and sky) added to the original image in Photo RAW.