We’ve all been down the road of conforming where the repos simply don’t match, even after doing a scale fix from dailies res to working res. I’ve spent hours diff matching plates and setting repos.
Autodesk, if you’re listening, I think it would be awesome if there was a button for the axis in action to snap to a reference. Essentially, I park on a frame in my conform (or batch), I have an action soft effect on my beauty plate and am looking at my reference on my secodary track. If I click the “snap” button (or whatever we call it), the axis would look at the ref and set the axis translation, rotation, scale to match. Maybe it would be a keyframe so you can do animated repos.
FYI our @FriendsFromAutodesk friends are listening, but, this is not an officially Autodesk moderated forum. Those that are here are here for the love of you and Flame and all that is our Logik community. For any feature requests, you will need to provide a Flame Feedback Request by following the procedure below.
And, to be clear, this has been the case for flame-news, Facebook Logik, and now, forum.logik.tv. If you submit the feedback request and post a link to the request here, I’ll upvote it with ya.
I’m a huge fan of feature requests, so this is great. It’s exciting when Flame gets new features, and even more exciting when we’re involved with them! Even in the context of these forums, although it isn’t an official Autodesk communication channel, it can still be fun and valuable to chat about proposed features like this with other Flame Artists. At least I think so! So since you’ve already posted here, why don’t we open the dialogue.
If I’m correctly understanding what you mean by “snap” in this context, I would be worried it might be quite a bit more complex than the kinds of snapping I’ve seen in other parts of Flame (first thing that comes to mind is the Adjust Offset button in action). I think the end result of what you’re asking for is full-fledged Machine Learning Difference-Based Repositioning. That’s a feature I just made up, but I haven’t looked into it; maybe the concept is out there! Granted, we’re getting close! Every release of Flame we get more and more machine learning features, so maybe this really is right around the corner. And you’re absolutely right that it sure would be a time saver and truly a game changer for anyone doing conforms.
But with this in mind, I’d be hard pressed to call it a snap when you need Flame to figure out how to transform, scale, and rotate the beauty plate with nothing more than pixels to work with. In other words, it really doesn’t have anything to snap to. The reference picture’s positioning of a clip is of course based on an editor’s decision. The reference picture doesn’t have any accessible transform information baked into it; it’s just a movie. An edl / aaf / xml on the other hand should have exactly that— transform data. But as you mentioned, we’ve all been there. And I‘m sure what you’re getting at here is that the edl / aaf / xml frequently doesn’t send transform data correctly due to any number of mysterious reasons: editorial programs communicating differently, dailies resolutions being different from the source resolutions, evil gremlins inside of Flame itself, etc— which ultimately leads us to spend a lot of time matching things.
Asking Autodesk to get better at reading an edl / aaf / xml is an age old issue with a ton of complicated variables. Maybe there’s hope there but it’s hard to say. Asking Autodesk to create a new feature that automagically matches a reference’s transform properties sounds to me like a conform artist’s dream come true and an incredible marketing opportunity for them. Let’s see what they have to say!
Yeah ‘snap’ is just a working term until something better is decided. Under the hood, I’d imagine it would use something like the shape tracker algorithm to match the plates. Should be an easy match since it should be identical frames (just different color, resolution and compression softness). I don’t think machine learning is necessary.