Behind the Scenes - On the Level Episode 2 (Patreon)

Hello! 

So, I just released the second episode of my interview series On the Level - the show where I play awesome video game levels, alongside the designers who made them. And because this one was such a mammoth undertaking, I wanted to share some behind the scenes info on how it came together.

After the launch of the Ori episode, I started talking to a few developers about future videos. An obvious candidate for me was Dishonored - the series is famous for its level design and there are about 4 or 5 stages across the franchise that would make for great episodes. 

After chatting to some people at Arkane, we decided that the Clockwork Mansion would be best. I scheduled a Zoom call with Dana and David, and then got to work on preparing.

Preparing for the interview

After downloading the game off Steam, I discovered that I still my have Dishonored 2 save files - which meant I could instantly access the mansion (unlike Will of the Wisps, where I had to play 70% of the game to get back to the Windswept Wastes). And so I just needed to play through the level a few times.

Unlike the Ori stage, the Clockwork Mansion is not a linear level - there are lots of routes through the house, different orders for the objectives, and tonnes of secret paths to take. And so I needed to come up with a specific route that would encompass all of the most interesting parts of the stage. I planned it out, beat by beat, and practiced my route ahead of time.

Then I could come up with some questions. I like to have relevant questions for the area I'm standing in, so the conversation flows nicely. And some general questions for other moments. I call them "panic questions", as they're something I can rely on if I don't have a relevant question on hand and I start to panic!

I ended up with an itinerary - a document I could have in front of me that reminded me how to tackle the mansion, and what questions I should be asking at each moment.

Conducting the interview

I do these interviews over Zoom on my iMac. My capture card - the Elgato HD 60 S+ - has a nifty feature where you can turn the HDMI input into a webcam feed, and so with a hit of a button I can switch my Zoom feed between a webcam recording of my face, and a live feed of the game running on my PC.

I find this is better than screen sharing, as the framerate is a lot better and its easier for the recording.

I record the Zoom call locally on my iMac - this has the benefit (over cloud recording) of spitting out individual audio tracks for speakers. I also output my iMac display to another capture card - plugged into my laptop - so I can create a second version of the recording. I'm always terrified of the Zoom call breaking and not recording anything, so this is my backup.

I also record my playthrough of the game itself, also onto the iMac, as that will appear in the video.

As for the interview itself, I just use the itinerary from above - but obviously adapt on the fly. I come up with new questions and decide not to ask others, depending on how the conversation flows.

This time I ran into a slight problem. One of my interviewees, David, is not a natural English speaker and so his answers were quite difficult to understand. In the end I asked him if he would prefer to give his responses in French. I don't speak French so it would be a bit of a weird conversation, but I'd deal with translation later.

After I finish the level, I ask the interviewee to share their screen and show any behind-the-scenes materials they might have and talk about them.

I got lots from Dana and David, including prototype level designs, Maya renders of the level animations, flow charts, and more. I asked them to then send across the original files, where applicable, as they'll be a higher framerate.

Processing the recording

Okay, now it's time for the big bit. The interview was well over 2 hours long, all over the place, and featured lots of tangents and nonsense. So I've got to get to work on rearranging it into a coherent narrative, and on cutting it down to size.

The first thing I do is load everything into Premiere and sync it up. So that's the Zoom call recording and the backup recording, the game playthrough, and the separate audio feeds for David, Dana, and I. They've all got to be lined up perfectly.

I then export this video and upload the whole thing to YouTube. That's because YouTube has a really good automatic captions system - in a few days it will have generated subtitles for the whole video, which I can then download and import into Premiere.

It's really useful to have this - it makes it so much easier to scan through the video when you can see what is being said, and it also gives me the power to literally search through the video. If I later need that bit where Dana talks about prototypes, it will be easy to find with Ctrl+F.

I also exported the French sections and uploaded those to YouTube as well. I then found someone on the GMTK Discord who speaks French and English, and paid them to translate the subtitles into English for me. Très bien!

The next thing I do is split the conversation up into individual chunks - usually my question and the interviewee's answer.

As I'm doing this, I use Premiere's marker system to detail what this chunk is about, and what key things were discussed. This gives me a really useful overview of the call.

Next, I copy and paste those markers into Keynote (Apple's version of Powerpoint).

My goal is to rearrange the chunks and decide what order they should be in, what I should keep, and what should be left on the cutting room floor. It's possible to do this in Premiere but it's just easier to do it in Keynote where I can effortlessly swap the slides around, add notes, change colours, and so on.

The order I decided on is pretty similar to our conversation, but various parts are removed and some stuff is moved around. For example, Dana showed me some prototypes of the level at the end of our call - but this worked better at the beginning, when we were talking about how the idea first came together.

Now, with the order (generally) sorted in Keynote, I could go back into Premiere and move the chunks of conversation into the correct order. I used some blank frames as headers for the different sections, to help keep everything nice and orderly for me. 

At this point the video is still about 2 hours long, so needs much more work.

The next part has two purposes.

One is to trim the interview responses down as much as possible. I don't want the video to be two hours long - and also I want the interviews to be snappy and easy to watch.

I know from looking at YouTube analytics that people ditch videos as soon as interviews start. I think that's because an interview answer is usually quite slow, repetitive, full of "uhm"s, and so on - especially when compared to my tightly scripted and edited narration. So to help with this, I make the interview answers snappier by removing most of the "uhm"s, and cutting the response down to its absolute essence.

Now you have to be careful to avoid the, what I like to call, Owen Wilson in the MCU effect. You don't want the cuts to sound weird and disjointed. So it's a case of using careful cuts and volume transitions, a little bit of audio overlap, and sometimes just deciding that you can't make a certain cut. I don't mind a bit of staccato editing - that's normal on YouTube (the Vlog Brothers style, where there's a cut between every sentence) - but it can't sound completely unnatural.

Every now and again, I'll find that the interviewee response is a little unclear - they might say something that doesn't work outside of the context of the original call. Or I need to put two bits of the interview together and there's no natural link in the audio. I can usually flub this by finding little bits of dialogue elsewhere in the call (hooray for searching through the subtitles!). For example, I might just need to find my interviewee saying "but" or "and then" or "because" - and then paste it in elsewhere in the call.

I also re-record my questions at this stage. It feels a bit weird to do this but it works well - again, I can make them snappier. And I can make better transitions that perfectly suit the new narrative I have constructed, rather than the original Zoom call.

And finally, I decide on what I'm going to say in my narration bits - in this case, what I'm going to say before and after the interview, and also what I'm going to say in the little interstitials between sections of the mansion.

Making the video

I next record my narration and plop that in. And now I've basically just got an audio track and need to figure out what visual stuff actually appears in the video.

The way On the Level works is that I have a part where I'm playing through the level (in full screen) - and then a pop-up window over the top to highlight other things like the interview video, cinematic shots of the level, footage of other games, concept art, behind the scenes material, and more.

Originally, as seen in my Last of Us interview video, I was going to have the video call in a little window in the corner at all times. But I found that it was hard to know where to look - should I be focusing on the call or the game footage? It was better to just make that decision for the viewer.

Anyway. This part is pretty messy, but the key idea is figuring out when I'm going to show my playthrough and when I'm going to switch to a pop-over.

Something I did early was record a bunch of "hero shots". So I loaded up the Clockwork Mansion, turned on a cheat that makes me invisible to all enemies, and removed the HUD and head bob effects. And now I could make little cinematic shots of all the key parts of the level - the transformations, specific details, overviews of different rooms, and so on.

In Premiere, I cut them all up and put them in a bin, which makes them easy to drop into the timeline at key moments.

I also played some Dishonored 1 and Death of the Outsider to get the footage for when Dana's talking about those games. And loaded in the behind-the-scenes material at the relevant points. I also tracked down more concept art and recorded myself doing specific actions that get referenced in the video.

This is just a long and laborious process. I'm trying to find a good flow, trying to get a good balance of playthrough footage and pop-over content, and trying to make sure the viewer understands where I am in the level.

Also, around this time I realise I should make a few more changes to the order of the conversation. Perhaps delete a section that ends up dragging the video down, or moving stuff around, or putting something back in that I previously cut out.

Once the whole timeline is sorted, it's time to finalise things. There's very much a "can't go back (very easily)" moment where I have to lock everything in and decide that it's final.

So first I create a "nest" for the pop-over footage (essentially a video within the video, so I can apply effects to the whole thing). I put a frame around the video and then apply a bunch of preset effects to scale it down, have the footage wipe in and wipe out, and give it a drop shadow.

I also export the audio from the call - Dana and David's voice and my re-recorded answers - and put them in Audacity to amplify and compress the audio. This makes it sound much better when I put it back into Premiere.

And then the usual final things - music, sound effects, little tags in the corner with game names, stuff like that!

And that's pretty much the whole thing done. I'll need to go back and edit the original call down to make it suitable for Patrons, and make the thumbnail and whatnot. But hooray! All done!

Hope that was interesting. Amusingly, On the Level was conceived as a way to make videos more quickly - a much easier thing to make while I was working on bigger projects. But as you can see, it's just as hard (if not harder!) to make than any episode of GMTK. That's a whoopsie on my part.

But oh well, it's fun, different, and vital for the channel. Each episode is packed with interesting info about great games and useful game dev knowledge straight for the source. So even though the views are pretty low (painfully low on this episode), I think it's worth doing.

So… onto the next one!

Mark