I started making music videos for fun last year. At first I worked on them on my phone almost every day, but my day job got busy and I took a break. I have recently started making them again. Many of the tools have changed while I was away, and I am still getting used to them.
Bringing the same person into the next scene
I used to make almost all my video images with Midjourney. I could find a frame I liked, but keeping the same person or place in the next frame was hard. When the hair or clothes changed, a continuation of the song could suddenly look like a different story. Keeping characters and scenes together was a persistent struggle.
Since returning, I can try more cuts than before, and it has become a little easier to keep a character recognizable from one scene to the next. It does not work every time. Still, I can think about where that person might go next, beyond a single image. The image at the top brings together material from three recent videos. I hope to find a small sense of movement within each song.
The days I compare myself
I feel as though I come across more AI-made work from people who know a lot about visuals or have wonderful ideas. This is a hobby, and I know I do not need to compare myself with anyone. Even so, there are days when I think, “I could not do that.”
That made me wonder how I could keep enjoying the work. I am still feeling my way through editing and structure, and fast-moving video or elaborate effects are difficult for me. Making one cut at a time and placing them alongside the music feels like something I can keep trying. Perhaps I can learn a little more from there.
If the story could last a whole song
I have also started talking things through with ChatGPT and Codex. I can consider the order of cuts while looking at the lyrics and music, and try an animatic—a rough video that joins storyboard images to the track. It feels a little easier to think about the whole song than it used to. A year ago I was doing almost everything on my phone, so the change in tools still surprises me.
For ‘Crow’, I moved from reading the lyrics and music to an artboard and animatic. In ‘未定でいい’, I tried following places from the lyrics: a station, a café, a bookstore, and a night walk. Some scenes seem to connect; some movements still feel off to me. I am still finding my way of making them.
I wonder if I could think through a world and its characters, and let one small story continue from the first cut to the last. The story might be simple, or I might think too hard and wander a little off course. I would like to try a few videos and see what kind of approach suits me.
I am also talking with ChatGPT and Codex about a way to read the music and lyrics and think through a story. I do not know yet whether it will take shape, but I would like to share it if it does. For now, I would like to make the next song all the way to its final cut.
Thank you for reading.
← Back to the journal