WBPP vs Sirilic

You should also try PI's FBPP; it was designed specifically to handle large numbers of frames 😉

wbpp gets slower kindof not linearly as the number of subs grow.

I have dekstop i9 14900k with 128gb ram.

A normal 100-200 sub project is done in 30-60 minutes with everything configured for quality.

the squid project of my was about 900 subs that took a good 10 hours.

Tony Gondola · Apr 28, 2026, 12:36 PM

Update!

With the release of the new AMSP script by Cyril Richard there’s no longer any reason to use Sirilic. The new script works just like WBPP, taking multiple file sets, even from multiple nights and sorting all out. You don’t even need to use the usual Siril folder structure. I tested it yesterday against WBPP and it completed in 25% of the time with very slightly sharper results.

DeepSpaceAstro has a video with all the details:

https://www.youtube.com/watch?v=80LwMI-WEJE

Awesome that’s cool to hear! I will have to check that out. I was happy with the results I got from WBPP but maxing my CPU out for 12 hours straight concerns me haha

My projects are typically 500-1200 frames from an IMX455 sensor so a typical stacking session with WBPP can take anywhere from 4-8 hours. I never want to tie up my primary PC (a MacBook Pro) with that job so I bought a “cheap” Rizen 9 powered Minisforum NUC style PC specifically for that task. It’s fast, headless, and connected directly to my NAS system that receives data from my scopes. I have it networked so that I can turn it on/off and sign into it from wherever I happen to be. I simply sign in, load it up with a job, press “RUN”, and go about the rest of my life. Usually by the time I feel like I have more time to look at doing the rest of the processing, the data is all stacked and waiting for me. This set up has been life changing. It’s easy, it feels seamless, and I’ve almost always got stacked data sitting around waiting for more attention. The only “minor” problem (besides the cost) is that with my current internet service, the upload speed is unacceptably slow. The stacked data files are pretty large and if I shoot all 7 filters, it can take 4-5 hours to download it when I’m not home. I approach that problem by creating a raw RGB combination so that I have fewer files to transfer. This summer, I’ve got to try to try to address that problem. Hopefully I can find a symmetric fiber service that doesn’t cost an arm and leg.

I recognize that this isn’t an ideal solution for everyone, but if you are committed to processing a lot of data, it’s a solution worth considering regardless of how you stack your data. Offloading the really time intensive jobs to a dedicated computer works really well.

John

WBPP sounds incredibly slow, based on many people’s description in the past few years. I still don’t get why it is so slow. Does it try to plate-solve every sub? PI’s plate solving is indeed very slow. If it try to do it on every sub, no wonder the whole WBPP process can take very long. PI’s local normalization is also a slow process, especially for a slower or fewer-core computer. If WBPP includes local normalization as part of its standard configuration, then this may explain the slowness. I found the normalize scale gradient script much faster, but I doubt this can be integrated into WBPP and replace local normalization.

Personally, I only use WBPP for calibration. I do registration, local normalization, and integration separately. I only do plate solving when it is absolutely necessary (like before PCC/SPCC, or before mosaic by coordinates). Overall, I rarely needed more than 10 hours to stack a target, even for my 50-hr projects with many subs on a slower, old intel computer.