EbSynth Test #1

With the prospect arising of potentially learning EbSynth in my job, I decided to familiarise myself with the software and grasp the basics/ have a bit of a play around. I’ve been wanting to learn it ever since I became aware of it last year when the beta version was released and it’s even been sat in the downloads file on my computer since then. I did watch a few tutorials way back when, but it looked super complex and unfamiliar, so I decided against it until now.

I wished I had opened it sooner!

Spoiler alert: the results aren’t great, but for the short time I’ve spent using the program so far combined with what I have seen on YouTube and Instagram, I can tell that I’m going to have a lot of fun using it and it has massive potential. It’s also open source, meaning it’s absolutely free and you can download it for yourself HERE. This is not a sponsored post by the way(!) I genuinely am loving it so far.

So basically, EbSynth is a program (or software- do they mean the same thing?!) which allows you to create altered videos from essentially one keyframe via a texture synthesis algorithm. Pretty cool. I’m not here to really explain the ins and outs or terminology of what it can do (check out their website) for that as they can totes do it better than me, but I will explain my process and my thoughts about it etc. Also there’s loads of tutorials on YouTube about how to work it, so check those out as well!

To get EbSynth to work, you need a piece of footage you want to manipulate and then at least one altered keyframe, such as something painted or hand drawn, or whatever style or medium you want the finished video to look. It’s kind of like a Snapchat filter, but more hi-tech!

Frame from the footage

For EbSynth to be able to do its thing well, you need to pick a frame with as much readable information in, meaning you don’t want any objects to be hidden, mouth to be closed, eyes to be shut, something overlapping the character etc. My clip had a fair bit of movement in it and a head turn, so I actually decided to make 5 keyframes.

One of the keyframes

I did this because I wanted to give EbSynth as much information as possible for it to work as well as it can. As there was a head turn, it was important that I keyed both the profile view and the back of his head, or EbSynth wouldn’t have the information of what the back might look like, therefore the animation wouldn’t look as smooth. Or maybe even super glitchy and weird. I didn’t stay to find out! Perhaps I could have got away with less keys, but for someone who is so used to drawing every frame all the time, five out of 57 felt like pretty much nothing.

I sketched them out real quick in TVPaint with a bold line. I just wanted to see the parameters of the program in this first venture and then later I’ll do more experimentation with styles, colour etc etc.

Anyway, once you have the keys and the image sequence (note: sequence, not a moving image file) of your footage, then that’s where the fun can begin.

As said, totally new here, so it took a little trial and error to get a clip I was happy with. The interface of the program is pretty basic, but don’t let that fool you! The settings take a bit of getting used to. Anyway, for this final version, I found that it was better to output each keyframe sequence into separate folders and blend them myself in Premiere after, rather than let EbSynth blend it for you by locating them all to the same folder, which I did in the first go.

I mean just look at this frame it created:

Frightening! 

Also, apparently EbSynth doesn’t like bits going out of frame/ edge of frame or it doesn’t really know how to handle it. If I’d have known this prior to using this clip, I would have chosen a different piece of footage which didn’t touch the edge.

For a first test, I am happy with what I’ve achieved, though I definitely want to push it further and get more creative with styles and techniques. As said, this was a super quick test to see what I can do with it (or if I could even work it for that matter!), hence the clip not being the most exciting thing to have ever graced the internet!

I think I can implement it not only in an experimental way, but also in a practical way, once I get the hang of the software. This would mean (in my case) colouring and/ or shading my rotoscoped line work (as those are the things which seem to add on the most time). As always, keep an eye out on here for more developments in my rotoscopy stuff!

Here’s the final version:

And here’s a side by side, so you can see the comparison: