My 48 Hour Film Project Experience

In the previous article I wrote how I prepared for this year’s 48 Hour Film Project in Warsaw. Here I want to tell you what actually happened and how the situation unfolded from my point of view. However, before reading further, watch our short, so that I can safely write about it without spoiling it for you.

Setting up

The process of setting everything up was uneventful, everything went smoothly, perhaps with a very small exception of me not being able to setup Decklink Mini Monitor monitoring properly. But with persistence and systematic attempts I managed to work it out in Resolve, Fusion, Adobe Premiere and After Effects. It never hurts to be prepared.

It so happened, that Olek, the DP at whose place we were staying throughout the event, had asked me to work on a vfx shot with logo removal, so after I set myself up, I was already knee deep in work.

In the meantime, a very long discussions about the script ensued in the living room. I did not have an opportunity to participate in it to the extent that I wanted, plus I also fully expected everything to change in the morning, as had been the case last time. This first brain dump was just a warm-up.

Our luck of the draw gave us again the road movie or a mystery. The elements to be included were:

  • a dumbbell
  • a soldier named Stefan or Barbara Piszczałka
  • a phrase “repaint it” to be spoken
  • a recognizable image of Warsaw

Logo removal shot turned out to be a tougher one, and I finished work at 2 am. We were still not sure about the overall story. I decided to call it a day, drove back home, and got some sleep. I knew I would desperately need it within the next day or two.

Working on introduction animation in After Effects.

Working on introduction animation in After Effects.

A slow start

I arrived back around 11 am, and the crew was just finishing loading up all the gear. Konrad — the director — asked me, if I could make an opening shot where there will be a huge explosion happening among the Warsaw city-scape. Remembering how long it took me to work on vfx shots last year, I had some doubts, but the time was too short to persuade him otherwise, so we only settled on shooting in 4K and decided to add a camera movement later. I knew then that things had changed drastically since the time when I left off, but did not get any more information at that time. They would fill me in later.

48 HFP has very strict rules — you can only use what you create during the time the event is taking place. There are few exceptions, and I wanted to confirm what exactly would be possible. I called the event hotline and learnt that muzzle flashes, fire and explosions were fine, but smoke was not. Well, that was good news. I knew I could manage to generate artificial smoke with particle generators, but explosion would be a different thing. Thankfully, I did have a few in my stock footage libraries, and if things didn’t work out, I found two other libraries that had what I believed we would need.

But time passed, and the footage was not incoming. I received the script, read it once, then twice, and was not very enthusiastic about it. Took the time to setup exporting presets, bins, titles, and other similar stuff, fired up After Effects to create the text introduction, rendered it out, ate something that our lovely catering girls prepared, and waited.

Finally, the first scene was shot (the one with onions in the lab), and both footage and audio arrived at around 4 pm. I did copy and ingest the old-fashioned way, without using new Resolve’s clone tool. Of course, I copied the source material both on the internal hard drive, and on the backup SSD.

Syncing video and sound in Resolve is pretty simple.

Syncing video and sound in Resolve is pretty simple.

Getting into the flow

The process of syncing sound is pretty straightforward in Resolve. One, you can try to dump the sound and pictures together in a single bin, and then choose to sync them by waveforms. I had moderate hopes it would work, but unfortunately it didn’t. Well, here comes the manual process — open a waveform file in the audio monitor, open the picture in the source view, and then align both. Once you do, click the link button, and there we go. Remember not to open the files from file browser, because the link will be inactive. It took me a few times to figure that one out. I also used flags to mark shots which had synced audio (blue), without additional audio (yellow) and bad (red). The last action was saving the scene number in clip metadata.

Finally, I created a Smart Bin with all good video footage for a given scene — video with the scene number and no red flag present — and I was ready to roll.

Well, almost. I should have adjusted video clips audio channels right then and there, but I was a bit in a hurry and expected the syncing to work as in Premiere, where there is no need for this step. I started editing the scene in a separate timeline, and found out that only camera audio appeared. I wanted to get to editing though, to see what is working and what isn’t when it comes to the pictures that were shot, because I knew that things would be progressing quickly from now on. I decided I would sort this out later.

The pictures were interesting, the scene somehow came together, requiring only a little massage, and with time to spare I started working on the grade. We shot on Panasonic G5 in log, recorded internally. I expected I wouldn’t be able to push the footage too hard. Out of numerous available options I chose Arri Alexa LogC to Rec709 3D LUT as a decent starting point, and worked from there. Since we were going for a post-apocalyptic theme, I decided on the usual bleach bypass, but with a green tint in the shadows, white highlights, countered by maybe slight orange-ish midtones to give it a bit more distinction from the typical look. It worked surprisingly well, so I finished the grade with a desaturated, cold vignette with a hint of a blur. I liked the look, and was pleasantly surprised how much abuse the footage could take without breaking down. And I was even more happy to see my machine working without a glitch, playing graded HD in real-time.

Grading in Resolve is pure pleasure.

Grading in Resolve is pure pleasure.

Footage for the second scene (both heroes walking with a wheelbarrow) arrived around 7.30 pm, and I went through the whole copy, ingest, sync process again, feeling already a bit pressed for time. Konrad saw the cut of the lab onion scene and the grade, was happy about it. The team went on to shoot the remaining scenes.

The wheelbarrow scene was shoot in 4k, despite me asking not to do that. As expected, Resolve was not too happy about it. Knowing that proxy workflow itself is a bit of a hassle, I first turned to setup optimized media in Resolve to resize to HD. Unfortunately, on Windows there is not much choice codec-wise. DNxHD had to do the trick.

And thankfully it did. The encoding was very fast, mostly due to rather small amount of footage. Of course, it would be even better if this had been a background process, but it worked seamlessly, without a glitch. I only made sure that I was not using optimized media during the final render and was off to the races.

I cut the second scene and, since I didn’t have anything better to do, I finally started to look into audio. After changing the channel configuration for the synced clips to account for all channels I realized, that I have to redo the edit to get the proper sound on the timeline. Not much of a problem, though I did have a few L- and J-cuts, where audio overlaps previous or following video clip. This proved to be a bit of nuisance in Resolve, as it was not possible to keep all audio files linked together and not linked to video. I had to enter the trim mode and manually select trim points. Or manipulate just the video. One way or another, it was a bit of a hassle.

I redid the edit, but again postponed detailed audio work until after the edit was approved. See the pattern here?

Editing a rough cut. Working with multiple audio channels could be improved.

Editing a rough cut. Working with multiple audio channels could be improved.

Full throttle

I did have some time to delve more into the script, and decided I still did not get it, there were a few things which didn’t hold up for me. I also started feeling the pressure even more. It was already 9.30 pm, and we still had quite a few things to shoot – at least according to the script. I was starting to worry that we might not have time to wrap everything up. Anxiousness that I hadn’t felt last year, something that was beyond my control. It also meant that it wasn’t my responsibility, but it would become mine on Sunday, before the deadline. Because in the end I was expected to deliver the end result.

At 11 pm the team was back with the dumbbell scene. We talked with Konrad, watched the cuts, and I voiced my worries about the script and time we had left. While they ate dinner, and discussed what to do next, I kept editing. At some point we showed the result to the rest of the team, to get some feedback. There were complaints about audio (unsurprising) but in general, people liked the direction and shape it was taking. Olek mentioned that maybe we could zoom in during the first scene taking advantage of the 4k shots. In retrospect it was obvious, and I followed his advice later. Even though the scene did work as a single wide shot, there were a few moments that I later trimmed, which made it flow even better. We also definitely needed some sort of a transition between the wheelbarrow and the onions. It was added to the list of takes for tomorrow.

The dumbbell scene had more takes, and took me more time to find the most interesting ones and choose which angles worked best. In the meantime, the script was being reduced to its bare minimum, we were still discussing how to end the whole thing. My participation was brief and intermittent – while editing, I was overhearing the whole discussion. When I felt strongly about some idea, I jumped into the other room, expressed my opinion, and then went back to work. The brainstorming was taking much more time than last year.

Around midnight the team left to shoot the bonfire, and I was still polishing the dumbbell scene. With it done I made the second pass on the wheelbarrow. Audio was still not doing that great. However, after watching the whole cut to date, I figured out that the end of the dumbbell scene – the soldier repeating his gun – might as well be an acceptable ending of the whole movie. This made my editing and script anxiousness go away a bit, especially since I thought we did have some good acting and interesting shots in the mix.

I was still worried about the establishing shot with an explosion in Warsaw. I couldn’t do much about it, and at some point around 2 am I decided I was done, and thought I should get some sleep.

Resolve 14 brings truly NLE-quality editing to this package.

Resolve 14 brings truly NLE-quality editing to this package.

Growing restless

I could not fall asleep, despite things being calm around. For two hours I kept turning in my sleeping bag, but no rest came. Stress and adrenaline were having this effect on my body. Well, at least I calmed down a bit.

At 5 am the team returned with the bonfire scene. I thought I would only copy the footage and get back to sleep. However, we ended up watching the cut of the dumbbell scene, then watching everything again, and discussing the ending. I asked for the establishing shot to be delivered as soon as possible, because I was worried, that I would be unable to comp it on time.

I edited the bonfire scene, we watched it again, and at 7.30 am I felt that it’s the last moment to get some sleep before the Sunday craze, especially since Piotr, our music composer, asked for the final edit around 10-11 am. I knew it was not possible at this time, but I exported whatever we had so that he had something to work with. Despite arising anxiousness, I managed to fall to sleep almost immediately.

Comping in Fusion. For such quick shots I can't decide if I like it better than After Effects. But it did the job, when AE failed.

Comping in Fusion. For such quick shots I can’t decide if I like it better than After Effects. But it did the job, when AE failed.

Going postal

The alarm clock wakes me up at 10 am. The craze is already beginning. The team heads off at 11. I try to move the edit from timeline-per-scene to a single unified timeline. I attempt to use Decompose in Place, but the second sequence overwrites the audio from the first. Must be a bug. I need to go copy/paste route. Switching times between sequences start to irk me. I’m finally done, I will be able to work on audio properly now.

I get the establishing shot via WeTransfer around 12:30 thanks to an extremely fast Internet and sensible mobile data plans. This has priority now.

I open After Effects, and start comping in there. After 30 minutes I reach the point where I have to paint on the original footage, and every time I try, AE crashes. This is a show-stopper. I waste another 15 minutes trying to troubleshoot, nothing comes out of it. I decide to jump ship, and move to Fusion. I hope I know it well enough to avoid pitfalls.

Fusion doesn’t want to read the codec of the explosion clips. Splendid. Adobe Media Encoder to the rescue, converted these to DNxHD. Alpha channel was lost, damn it. I do a rough comp using luma keyer, the result is not satisfactory to my eye. I try to comp smoke from another shot. Everything goes smoothly up to the point where I need to add the Color Corrector node to better fit it into the scene. Another repeatable crash as soon as I click on the color wheel. No time, need to work around it. Use another node, maybe Brightness and Contrast, different values for each channel. Took more time, but around 2 pm we have something that does not look atrocious. Does not look that good either, but maybe nobody will notice, especially with the heavy grade. And Piotr is waiting for the finished cut, so I need to move on. I render the comp out in OpenEXR, put it on the timeline, grade it.

The team arrives with the rest of the footage around 3 pm. There’s not much there, just the final scene before the execution, the onion being planted and the transition between wheelbarrow and onion lab scenes. I have to sync sound first. Andrzej was not available for today, somebody else was handling the recording. The file names are off. It’s also a different recorder and a different channel setup altogether. There are no hand claps (no time for a slate ever) recorded, I am exhausted, and syncing takes more time. Out of frustration I shout at the sound man. It’s not good. I almost never shout.

I take a deep breath and decide to take a break for five minutes. I start my laptop to encode explosions into ProRes 4444 to save the alpha channel. In the meantime, I cut the final scene and the transition. We watch the whole thing with Konrad and the team. It’s good, except for the establishing shot (though I am the only one who minds it). We decide to shorten the bonfire scene to remove some chatter. There are other things that we could have done to make it shine, but there is no time. Discussions with the director, decisions on what we can and cannot do.

The explosion footage is converted, I give the establishing shot comp one more go, limiting my involvement to 10 minutes. It works. Looks a thousand times better than before. I export it again.

The final cut is ready at 4.45. Exporting a preview takes 5 minutes. Piotr gets the final cut to polish his work, I start working on audio. I realize there is no dynamics on buses in Resolve 14. I realize I cannot chain buses together into another bus. I realize there is no separate dynamics plug-in in Resolve. I find out there is no pan effect on clips, only automation. I know I have to be ready to export the final version at 6.30 pm and that I don’t have time to move the project to Audition. I stop responding to people’s requests if they are not vital, ask them not to disturb me. I realize my headphones do not cut me off as well as I hoped they would. Speakers are positioned too low to get proper monitoring.

I realize I won’t make it. All hell breaks loose in my mind. But I don’t have time for that either.

I manage to sort out audio channels and remove audio elements that interfere with each other. Just lav mikes and the boom where appropriate. I drop in the bonfire that Andrzej – our sound guy – recorded at night. There is only time for a rough volume treatment. I make one pass at pan, but Resolve crashes several times on me when I move between Edit and Fairlight rooms. When I open the project I do not realize that the automation is gone. I consider mono mix, but I can’t make it working properly under pressure. All right, the thing is audible, has to do. Hopefully Andrzej doesn’t kill me when he sees it. He has done a great job, and now I am screwing it up.

Export audio to make sure nothing else breaks during the final render. Put it in the output sequence underneath the video, make sure it’s synced. Add music. Adjust music levels slightly. Done.

6.30 – we watch the whole thing for the last time. Last minute level adjustments on the exported audio file, some changes to music – remove some, copy from the other place. No time. Render. Bless Quadro K5200 and my workstation, that thing is fast.

6.50 – the final render is ready. Watch it again. Render is fine. Copy it to the USB stick provided by the organizer. Damn, why can’t it go faster? They saved money, bought USB 2.0 pendrives, copying takes 10 minutes instead of 1.

At 7.10 pm Konrad with somebody else rush off to drop-off. They are on time. We qualify. Not everyone was able to make it. Relieved. Exhausted. Everybody celebrates.

I had problems with Fairlight tab, and it can definitely be improved a bit to reach the functionality of Audition, but it is already a powerful mixer.

I had some issues with Fairlight tab, and it can definitely be improved a bit to reach the functionality of Audition, but it is already a powerful mixer. The picture shows version 14.0.1 with Dynamics already added to buses and master channel and fixed automation bug.

Winding down

I am not sure what I should be doing next. Last year we made some changes after the deadline and delviered the final render at about 9 pm. We had to, because there was a rendering glitch in Resolve with ProRes clips, and red plate saying “media offline” doesn’t look good on a big screen. The organizers were kind enough to allow for it. This year I am too tired to keep working, especially since everybody else has also relaxed. I am out of ideas and my decision power is spent. I loiter. The sound will remain unfinished today. I start packing up around 8 pm, when it seems that I won’t be able to do anything else anyway. It’s been a tough ride. Thanks to my earlier preparations, this time I am not missing anything from my gear.

Konrad has birthday on Monday, so we throw a small party when he gets back from drop-off. Watch the final movie again with the team, everybody is impressed. Andrzej doesn’t kill me, though later critiques the final result. I accept and agree with everything he says. We hope to fix it before going public with the final version.

I stay until 10 pm and then head home. Driving a car at night calms me down, but I still feel excited. I am extremely happy about that movie. I pulled it off, regardless of the shortcomings. It has been the most satisfying thing I’ve done throughout my whole post-production career.


Addendum I: The Aftermath

You probably already guessed that I did cheat a bit. What you saw at the beginning of this article was the version that we did massage after the festival. If you want to see what we actually submitted, here it is (warning: no subtitles):

Later next week I saw it projected in cinema, and it held up pretty well, despite raised black levels (exported as auto, hoping for the best) and not so great sound. And of course hideous compression artifacts, a killer to all dark images.

Our hard work brought us four awards: Best Prop Use (the dumbbell), Best Costumes, Best Make-up, and Best Actor (for Waldek Nowak, the older hero). I am extremely happy for these – all well deserved. Of course, I wish there was something post-related among them, but it only shows that the competition was tough, and that I still have things to learn and room to improve. Maybe next time.

Addendum II: Lessons Learnt

Without question the biggest lesson for me was that avoidance tactics is equal to setting myself up for failure. It’s fine to be anxious, but the way to deal with it is to tackle it as soon as possible, instead of waiting until the last moment. Problems are not going to disappear. On the contrary, they are going to become much worse once stress and pressure are involved. Learning to notice the signs of such dismissals (such as “I’ll deal with it later”) is crucial. Not easy, but this is what needs to be done. Running away is not going to help.

Therefore in the future I want to approach several things differently. First, I am going to ask for any vfx plate to be shot as soon as possible, so that I can finish it earlier, when there is still more time to account for possible mishaps. Second – which actually should be the first – I will make sure that I have good sound monitoring and all tools that I need to do the acceptable sound in my NLE. I have already made some preparations in this department, getting some of the Waves plugins during their 25 year sale, and Focusrite Scarlet 6i6 interface that comes with a bunch as well. But most important of all – I will not keep pushing things that I feel not comfortable with for later, and when I feel myself taking the easier path, I am will take a step back and reconsider.

Third, workflow is always something to optimize. Being set in the old ways of “audio after picture lock” does not work anymore, especially since these days we can do so much in the NLE itself. Being able to avoid conforming and tool switching was vital to the success of this project. This needs to be maintained and explored further. That said, I am not yet sure how to modify the audio workflow. I prefer to edit scenes in separate timelines. My initial thoughts are that I should obviously get all the channels on the timeline from the get go, and work on volume and channel selection as part of the edit. Maybe massage individual files as well, if need be, though Resolve does not give me this opportunity – yet? Automation and track setup is not possible to easily copy over to another sequence, but maybe if I do the EQ/Dynamics once, redoing it will be easier than avoiding experimentation until the very end. Of course, ultimately nothing beats having a qualified sound engineer – as you can hear in the final version. And yet, with the time that we had, conforming and exporting, dealing with possible sync issues, could have literally made a difference between making and not making it. For this type of projects, a Swiss-army knife that combines editing, grading and audio seems to be the way to go.

Finally, it always pays to know the limitations and capabilities of the software. But that can only be gained through practice. Did my decision to go all-Resolve helped me or hindered in the end? It’s a good question to which I do not have an answer. Absent time travel, this experience is impossible to repeat. I can safely say though, that at no time during the editing process I regretted the decision or felt the urge to switch to Premiere. On the contrary, it’s been great to have been able to put Resolve through its paces. I think it passed the test with flying colors.

Tagged , , , , . Bookmark the permalink.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.