If any one with a RTX 4090 who could test TPAI with Denoise beta strong (5/5), Sharpen Strong ( 5) and Enhance (10/10/10) enabled at the same time .
On 5 images in Batch, just tell me the time it did take after you did press start, a stopwatch does the job.
That would be really nice.
Im working on a Semi Automatic Batch Processing with DxO Deep Prime, Capture One, Camera Raw, Photoshop (Nik Color Efex) and TPAI.
For images where time is key but Quality does play a role also.
This is for the selection of images where the customer chooses his images but afterwards the cropping and a slight retouching takes place.
CPU does play the major role here but i don’t know how long TPAI, with enhance enabled, will take time with a big GPU.
I did test the process itself but i didn’t do any benching, DxO DP XD does take 53 minutes alone with 800 images (Radeon Pro W6800), a 4090 will half this time with DxO.
The problem with this approach is you’ll never have a consistent outcome or workflow. For example, the “new” beta noise reductions are worse (for me) regarding results.
So, last week, noise reduction worked well for me. Next week it may not. That being said, the opposite can be true for someone else.
I’d rather see quarterly updates vs. weekly… with “Beta noise” being included for weeks at a time vs. a week or two. This way, we can alter our workflow over a period of time vs week to week. This would also avoid the debacle with the Lens/Motion sharpen models.
If the new models would be removed from TPAI i would move back to Denoise A.I., bc of the old Standard model.
With the Standard model i was able to finish my images after they where Processed with C1 and DxO DPXD and remove the last bit of noise.
I did test them also with other processes like denoising after C1 or LR and they were superior compared to the models in TL Denoise A.I.
I was also able to simulate a raw process with the new models and they where on par or better than Denoise A.I. of Adobe, since in raw processing you don’t need such high values like in processing images that were processed before denoising.
And with low values the new models work very well.
Don’t shoot me, but when it really counts, and I need predictable noise reduction without any of the current AI’s vices, I’m still going back to DeNoise 6.0.1.
