The Making of a Trailer
Today I released a trailer for my iPhone app Alcohology. This is the first trailer I've ever worked on1 and I documented the process. This post is a behind the scenes / making of / evolution / post-mortem rolled into one. It's got a little bit of everything.
Here's the final edit. If you'd like to follow the process in "real-time" then don't watch it until the end.
Why make a trailer to begin with? I launched in May and in the five months since I've sold 193 copies with zero coverage from any review site, blog, or publishing entity. This is part of my attempt to get noticed.
I'm only a programmer and there's no way I could make a good trailer by myself. Fortunately I work with incredibly talented people during my day job2 and they were willing to help out. All video editing was done by Steve Thompson and the original music was produced by Howard Mostrom. Lighting and camera work was done by my girlfriend Liz Heidner. Without these superstars the trailer would have never been possible.
The goal from the start was to demonstrate what this app could do and why it was cool. Once Steve was on board I told him three key points that we needed to get across.
- Users can customize ingredients and browse recipes that can be made using those ingredients.
- Demonstrate suggestions feature.
- Demonstrate random feature.
With a general idea in place we needed test footage to make an animatic. Armed with a Canon Powershot SD880 (~$300 in '08) point-n-click I grabbed all my liquor bottles and took as many shots3 as I could think of. Steve edited this into the following animatic. It was originally set to placeholder music which copyright law prevents me from including so there is no audio. :(
The big take away was that fast-motion drinking mixing is awesome, particularly for the Toucan4. The first 30 seconds of the trailer however are just kinda boring. The same test footage was re-edited to produce this.
Much better! This edit proved our concept and served as a baseline.
With a solid concept in place it was time for a more professional setup. This is where Liz came in. She's a hobbyist photographer with a keen eye and access to a Canon EOS 7D DSLR (~$1500 + lens) which is a massive upgrade. It's primarily for still shots, but takes great video at 1080p. Shooting at 1080p was great as it gave lot of room to crop an ideal 720p frame.
To say that our setup was ghetto would be an understatement. It required duct tape, wire, stacked books, Amazon Fresh crates, five household lamps, and several trips to the fabric store5.
The first filming pass with the Canon 7D focused only on the Toucan, bar setup, and lots of experimentation. Here is the next edit which contains a mixture of old and new footage.
The new footage looked amazing, but there were numerous production issues. Device angles (:17), overblown lighting (:28), poor framing (:34), and jerky motion (:36) amongst others.
At this point we filmed the Toucan over and over and over. Different setups, different framing, alternating pours left/right vs all right, empty shot glasses in frame/out of frame, and more.
After awhile we hit our first major issue. The setup and camera worked great for still shots, but was horrible for any handheld motion. The lens of the Canon 7D is fairly long causing it to magnify any shakes or jerks. Just keeping the target centered proved impossible by hand. Here is an amusing clip from Steve trying to demonstrate proper motion.
We devised a partial solution that allowed for basic sweeps and pans. I present to you the ghetto dolly.
A piece of cardboard with a tripod on top! Ok, that's a bar stool because I took the picture after filming when the tripod was unavailable, but you get the idea. The camera was set to record while I sat on the floor and slowly pulled and then pushed the cardboard across the floor. A second take was done with Liz turning the camera to stay centered on the target. I hate to say it, but this actually worked. It required a few takes to minimize shake, but it worked.
Unfortunately it did not enable elaborate handheld shots. The final trailer only has advanced camera maneuvers for the Toucan. Steve achieved this magic by setting the camera crop frame by frame. Sorry Steve!
Music and Audio
While iterating on the video edit Howard was working on an audio track. It was initially made to the approximate time of the temp track and early edits. Steve then made minor timing adjustments to hit the right notes so Howard could finish it off. Here's an early look at the audio.
The goal for a trailer is to get people excited for a product and this didn't do it as well as the placeholder. It was a little too jazz lounge like, particularly around the first and last mixing shots. Here's some experimental audio to try and spruce things up a bit.
The old radio voice and crowd noise were deemed too distracting. It does however hit much harder at key moments which is what we wanted. The final audio is a fine tuned version with the voice and crowd removed.
If you chose to follow the trailer in "real time" you may now go back to the article start and watch the final edit. :)
Trailer 2: Trail Harder
If I were to do this all over again there are a few things I'd do differently. First and foremost is experiment with an actual camcorder. The Canon 7D DSLR had awesome quality, but the inability to perform advanced movements sucks. The camera is simply not intended to be used in that way. I have no idea what the quality of a camcorder would be, but I'd love to find out. Alternatively you can rent Steadicam rigs starting at $75 a day or do-it-yourself for $30 (clicky). It'd probably still suck with the DSLR.
I'd also want to decrease filming iteration time. A single take on mixing a drink could take as long as 30 minutes. Planning, prepping, filming, and cleanup was ~15 minutes. Yanking the memory card, copying to PC, viewing, and discussing was another ~15. Having a live monitor setup attached to the camera where the footage could be instantly viewed would have been rather nice.
You'd think it wouldn't take long to record footage for a 70 second trailer. Protip: It does. All of the footage for the final edit came from a marathon 12-hour Saturday. That was only possible after ~4 nights of 2-3 hour test sessions. It's a lot of work.
There were about 7 passes (film, edit, discuss) made for the video plus 3 or 4 for audio. Raw footage was transported on thumb drives6 and trailer edits were shared via public dropbox folders. The whole process took five weeks, but that is in large part to this being an after hours project for everyone involved.
Working on this trailer was a huge learning lesson and a lot of fun. No matter what happens next I'm proud of how it turned out and have no regrets. Countless thanks to those who assisted me and I hope this post is informative or helpful to you folks on the intertubes. Cheers.
Time: Planning and Discussion, 4 hours. Filming, 22 hours. Editing, 25 hours. Audio, 16 hours.
Hardware: Canon SD880, Canon 7D
Software: Sony Vegas, video edit and post. Photoshop, 2d work. 3D Studio Max, 3d intro work. After Effects, vfx and post. Digital Performer, composing. Dropbox, file sharing. Protools/Sound Miner/Peak, mixing and sound effects.
- I have observed the production of many trailers at work, but was never a participant.
- Programmer at Uber Entertainment working on Super Monday Night Combat.
- Film shots, not liquor shots.
- Thanks to Brittany Aubert for sharing this eye catching recipe.
- If you're a dude who walks into a fabric store you are immediately assisted because you obviously have no clue wtf you're doing.
- The final push was 12gb of video data and required 3 thumbsticks!