NLE best suited for hosting Rawtherapee as raw engine?

Hello, I am wondering if anyone has any ideas on what video editor it would be most easy to program a plugin that uses Rawtherapee’s core features as a third party video raw engine?

I found that I have unparalleled raw editing control, with everything from demosaicing to Flatfield and dark frames in Rawtherapee, which is why I initially started using it for processing raw video, but am trying to find a better workflow than batch export in tiff using the command line. My old tiff workflow is workable, but it means I don’t have the options of setting things like white balance and exposure in a proper video color grading environment, where I can easily scrub through a clip, toggle between clips and times for easy color matching, and keyframing base parameters using the NLE GUI.

My current workflow is time consuming, from working to fit all the dynamic range in a 16 bit tiff without messing up colors and tonal response, to raw processing all the frames of a clip that only snippets are used. Of course, there will always be small ways to improve the workflow options I currently have, but having Rawtherapee internally in a video editor would just make everything easier, and I could just let things fall at whatever exposure they may in the raw processing stage, not worrying because all program/plugin/program pipelines are floating point.

Perhaps a huger benefit would be that Rawtherapee video processing would become accessible for people who do not have the will or patience for command line processing, and would just become more accessible, overall.

My plan would be as follows: Find what NLE is best suited for this and get familiar with it, then learn what programming languages and APIs are necessary to make it work, and then find someone who can do it or learn how to do it myself.

I currently only have limited programming knowledge, just doing light command line actions, and making basic Python for loops and calculations. See this link, GIMP acronym recursion , as an example. After hearing the story of aurelienpierre self - teaching C++ in a few months and then getting in the hot seat making major improvements to Darktable, I feel that I could learn to do it or recruit a friend with a stronger CS background than I to tackle this.

Also, if anyone with moderating experience has an idea for a better subforum to to move this topic to, I am fine doing a copy, paste and delete. This just seemed the most logical place for it to me.

If you’re looking for free and open source solutions, then I think that KDEnLive is the best. @frd is very involved, so perhaps he can shed some light on things.

I’ve also heard good things about Olive and shortcut, but I haven’t used them, and I’m a bad video editor anyway.

I am open to FOSS or proprietary NLEs, just something that is well developed and stable.

Hello,

Kdenlive uses MLT as its framework to process video, so you’d have to develop something at that level. Would be best to send an email to the MLT team mailing list to see if it would be feasible and how to proceed.

Cheers

1 Like

If he’s exporting TIFFs, he’s likely wanting the NLE to work at higher bit depths than 8 - which sadly eliminates nearly all opensource NLEs, because nearly all of them use MLT and so far, and 10-bit support is extremely low on Dan’s priorities list - https://forum.kde.org/viewtopic.php?f=265&t=112573&p=270839&hilit=mlt+10+bit#p270839 - that’s and old post but nothing has really changed since then.

Unfortunately, fixing the core engine to no longer operate on int8 is a massive undertaking so I can see why it still hasn’t been done. I’ve periodically tried to sift through MLT’s code in order to try and fix some issues I’ve had with kdenlive (it’s notorious for behaving poorly with full-luma-range video) but it’s architectured in a way that I found very difficult to follow.

Yes, I guess in 2009 it wasn’t very important for him. From when I last read him talk about this it seems that he might be willing to change this only after the release of Qt6.

Mlt has Movit which uses 16-bit float textures, which can support a 10-bit integer input and output. A huge stabilization effort was recently made and there seems to be only 1 issue remaning (more testing needed of course).

Improving Movit would be a better solution than rewriting everything from scratch. The Kdenlive team is trying to look for a solution, maybe a crowdfund, but there isn’t anyone with such knowledge interested in making it happen. Not an easy task…

1 Like