I have both a Seestar S50 and Dwarf 3 and I always run them side by side every night. I want to discuss a comparison between the on app image postprocessing, what ZWO calls AI Denoise and DwarfLabs calls Stellar Studio.
My goal here is not to say which one is better, this is not the right place for that. If you really want an answer, they both fall short of what could be done, but in different ways. What I want to do is point out the few things Stellar Studio does better in the hope that ZWO listens and does something about it.
AI Denoise is local, it runs on your phone actually. It downloads the captured image over WiFi from the S50, hopefully in FIT not JPEG format, although it is not clear if that's the case. It only does one thing, and is relatively fast, about 15 seconds for a 1x image and 60 seconds for a 2x mosaic. The results are OK, but its main drawbacks are the lack of a dial to control the amount of processing, which tends to be super aggressive and more importantly the fact that it only saves in JPEG format. Both should be very easy to fix.
Stellar Studio is a very different thing. It runs in the cloud, you transfer your image from the telescope to the phone, then you upload it to some server in China. This takes a long time, with the uploading step especially problematic, sometimes I have to try 4 or 5 times until it succeeds. It can take anything from a minute to half an hour, depending on how lucky you are. Once you managed to upload your FIR file, there are 4 things you can do, Auto, Star Correction, Denoise and Star Removal. Auto does mainly background extraction and maybe some luminosity and contrast adjustment. This is also where the mosaic assembly happens, unlike the ZWO mosaic process that occurs in parallel with live stacking, the Dwarf simply takes four images in 1.8x mosaic (the largest possible, but its filed of view is also almost twice the S50 one). This mosaic stitching happens now and it is far from perfect, the overlaps of the four sub-images are clearly visible. In this respect, ZWO is way better.
The other three steps do what they say they do, some star deconvolution, noise removal (this more or less matches ZWO's AI Denoise) and interestingly, star removal. But while you get a starless image, you do not get the corresponding stars only one, so then what's the point? The idea is that you post-process them separately and then merge them back. The best you can do is subtract the starless image form the denoised one yourself, it sort of works but it's a hack. Finally, you can generate an overlay/watermark, but instead of creating a separate file, you can only superpose it over an existing image, so you have to subtract images again to get a clean overlay layer image.
So you have a lot more options, but they are slow, each one takes between one and three minutes. The other good thing is that you can save JPG, PNG or FIT format files at every step and the uncompressed ones are quite large, up to 100MB if they are 1.8x mosaics. Transferring these files from the cloud is not fast and everything is manual and cumbersome. It can take up to 30 minutes spent phone in hand to go through everything for a single file - this brings back unpleasant memories of ZWO's Seestar a year ago. Post processing all the images taken during one night with Stellar Studio is a multi-hour endeavor.
In conclusion, Dwarf Lab's Stellar Studio does a lot more than ZWO's AI Denoise, like multiple ways to improve the image and saving in PNG and FIT format. It is however inferior when it comes to mosaic stitching, much slower and more cumbersome to use. If ZWO wants to maintain the lead, they should seriously consider improving the image post-processing, which for a consumer level product is as important, if not more, than the imaging itself.
