storytelling with 2D animation

I made a short anime intro video of myself without opening a video editor, just by giving notes to an agent, and it took eleven versions to get it right..

So last week I made a short intro video of myself.. well, an anime version of me, talking in my own voice, and I didn’t open a video editor even once.

Play the intro video: my anime self on a terracotta background with speed lines and the letters KV.

Play the video

I already wrote about getting back to building, so I won’t go into the tools again. This one is more about what it felt like to direct an animated video when all I had was notes..

It started with my character sheets. I gave an image model a few photos of me and asked for an anime version in my colours, and it got a lot right but also some things wrong, like the tattoo on the wrong arm and chains I don’t wear, so it took a few rounds. By the end I had nineteen sheets with gestures, walking, jumping, exercise, yoga, head turns and even a mouth chart for lip sync.

Then the agent cut those sheets into about ninety drawings and built a small engine that plays them pose by pose, the way animators do, and adds breathing, blinking and the hair moving on top of that..

A cropped corner of the Jump Sequence character sheet: three drawings of my anime self mid-jump, fists up, with a terracotta box around the middle one. — A corner of one of my nineteen character sheets.

Six of the cut-out drawings on a dark background, labelled g00, g01, g05, j6, jog1 and zen: waving, pointing up, hands up, mid-jump, jogging and sitting cross-legged. The jump is underlined in terracotta. — A few of the ninety or so drawings cut from them.

A screenshot of kv-choreo.js in a dark editor: lines that pick which drawing plays on which spoken word, with the jump line framed in terracotta. — The code that decides which drawing plays on which word.

A frame from the making-of video: the waveform of my cloned voice with a terracotta playhead, each spoken word placed on its own spot in time, and the word CUT below. — My voice as the clock – every word has its own spot in time.

The voice is a clone of mine. It kept saying Kalyan the wrong way, so in the video I’m KV, and for the line with my website it uses my own recorded voice. I think the part I liked the most is that the voice is the clock.. every word has a timestamp, so the cuts, the gestures and the captions all land on the words, and if I change the script the whole video retimes itself.

A diagram of how the video gets made: my script goes to my voice clone, which gives every word a timestamp; the timestamps drive the scenes, where the cuts, gestures and captions land on the words, along with about 90 drawings cut from my 19 character sheets; every version is rendered and checked, I watch it, and one line of notes goes back for the next version.

So the directing part.. the first version looked forced, and that’s exactly what I told it. It measured why before changing anything, the frame was standing still for almost ten of the thirty seconds, and it fixed that first.

Most versions moved like that, on one line from me. “My body isn’t moving”, “the left hand looks dislocated”, “it’s 10+ years, not 8, and I’m still a product person”, “the rewind looks buggy, not intentional”.. and every one of those changed something in the next version.

Four versions of the same scene side by side: v2 and v6 say 8 years, v8 and v11 say 10+ years with a board of product work, and v11, the final, is framed in terracotta. — The same scene in four versions. Eight years became 10+, and the labels on top went away.

At one point I said it all felt more childish than credible, so it made a calmer version, and honestly I didn’t like that one. I liked the energy of the earlier version, what I wanted fixed were the bugs and the wrong facts.. so version ten went back to the energetic one with every fix in it, and version eleven took off the last labels I didn’t want.

So a few things I’m taking away from this.. first, my best notes were about how it felt to me, like “looks forced”, and those did more than a long list of instructions would have. Second, getting the small facts right mattered more to me than polish, eight years instead of ten would have bothered me more than any glitch.

Third, slowing it down helped, a little less than a second more between thoughts and the same story was so much easier to follow. And also I could check everything, every version came with frames, sound levels and even proof that all nineteen sheets were on screen, so I never had to just trust it..

All of it sits in one place now, the character, my voice, the scenes and the steps to make the next one, so my next video is basically a script.. the first one I made with it is a thirty second follow-up on how this one came together.

Play the making-of video: my anime self under a terminal command and the words NO EDITOR.

Play the video

The old constraints kept me from starting.

Now I can go from an idea to a finished video in my own voice, and that keeps me from stopping.

I’ll take that trade.

storytelling with 2D animation

Discussion