Akira Kurosawa - Composing Movement

Part of this, I think, is also the shift to dialogue-based scripts. It reminds me of this excellent Every Frame a Painting video that compares Kurosawa’s effective use of movement with the barely-moving, dialogue-driven Avengers films. A huge percentage of films/shows today are what Hitchcock called, “photographs of people talking” and aren’t actually cinematic. If they were more cinematic, hearing every line of dialogue wouldn’t be as important to understanding the story. Start at about 4:25: ▶ this video

I think we're sort of aggressively agreeing with each other. It's not the CGI (or lack thereof) that makes the modern superhero films bad (except as a sort of demoscene-with-infinite-budget visual spectacle where they're great). It's the writing. To a lesser extent the cinematography is poor, which makes it hard to really feel what's going on. There's lots of unnecessary camera movement, frenetic jump cuts during fights, and all the colors are dull and muted. Every Frame a Painting has a good video on Kurosawa's use of movement in his films, and actually uses The Avengers as a contrasting example and how the cinematography & direction makes the scene in question feel meaningless.[1] So I agree that it's not the CGI that's the problem. It's the writing, the direction, the cinematography, the editing, the music[2] (though they improved since the linked video came out, directly in response to it), etc. The films are bad in many ways, but the CGI isn't really one of them. [1] ▶ this video [2] link

As an artist and designer (with admittedly limited AI experience), where I feel AI to be lacking is in its poverty of support for formal descriptors. Content descriptors such as 'dog wearing a hat' are a mostly solved problem. Support for simple formal descriptors such as basic color terms and background/foreground are ok, but things like 'global contrast' (as opposed to foreground background contrast), 'negative shape', 'overlap', 'saturation contrast' etc etc... all these leave the AI models I have played with scratching their heads. I like how Veo supports camera moves, though I wonder if it clearly recognizes the difference between 'in-camera motion' and 'camera motion' and also things like 'global motion' (e.g. the motion of rain, snow etc). Obligatory link to Every Frame a Painting, where he talks about motion in Kurosawa: ▶ this video The abiding issue is that artists (animators, filmmakers etc) have not done an effective job at formalising these attributes or even naming them consistently. Every Frame a Painting does a good job but even he has a tendency to hand wave these attributes.

What I love about Every Frame a Painting is the depth of its research. It's quite clear that the creators had an extensive knowledge of film. The videos always include wonderfully selected interviews with the filmmakers that emphasize the point far more than just a person talking. The Deakins interview in Shot | Reverse Shot taken from Cinematographer Style is a nice example^[1]. Or Sidney Lumet's interview in the Kurosawa video^[2]. Taking from film criticism solidifies the video's arguments, creates an amazing rhythm to the videos (Tony and Taylor are clearly great editors) and also prevents the "annoying person endlessly lecturing to me about film" feeling that I get with some other channels. I've thought about doing videos of my own on various topics (the usage of eyes and glances in Il Posto, Kieslowski's poetry, the reflexivity of In the Mood for Love, etc.) but the YouTube film criticism industry is pretty saturated. Perhaps a different format? [1]: link [2]: ▶ this video

I think the main issue with Prezi-style presentations is that it demands too much skill to use properly. A better interface with sensible defaults and much less emphasis on wow-factor would help a lot here. Normal powerpoint "only" requires a basic awareness of composition and narrative coherence, which already is too much to ask in practice. To make good Prezi presentation you need to have the skills of a cinematographer/editor. Specifically, the skill of continuity editing[0] so your audience can follow what is going on. What we want is the calm, easy to read compositions of, say, Yasujirō Ozu or Wes Anderson, and Akira Kurasowa's sense of movement[1][2]. Instead everyone uses it in a way that it feels more like the worst aspects of Bayhem[3]. I still believe Prezi-like presentation tools have the potential to be good, because they could be used to map relationships in your narrative to spatial relationships. But that would require a) that you are aware of those relationships yourself and b) that you posses the skills to translate that into spatial compositions that are coherent. This is obviously an unrealistic demand, so instead the tools should assist you with that. It needs is an interface that helps users structure their story, with sensible default conversions. Maybe something like Treesheets[4], where you can create a tree-based hierarchy in your topics, and then a conversion of that to a spatial hierarchy that is easy to follow, and not a distraction from the actual narrative. The vertical slides in reveal.js are a step in the right direction, starting from the powerpoint side of the spectrum[5]. What we need a step in the right direction of the free-flowing Prezi side of presentations. At some point there will be a happy middle ground. EDIT: Maybe we should all just work our way through Pixar's storytelling course on Khan Academy[6] [0] link [1] link [2] link [3] ▶ this video [4] link [5] link [6] link