DaVinci Resolve 21.1

The person you are replying to seems to have a depth of experience with editing over many years.

It seems pretty clear that AI capabilities are jagged. There are good reasons to believe based on research and the subjective evaluation of experts who are quite friendly to AI that there are some tasks creative tasks that AI has been really quite good at (totally disrupting stock illustrations) some where it does so much better than typical person, but not nearly as well as well as an expert (song creation) and some where, so far, it pretty much fails, and sees almost no progress even as models get better and better. (long form creative writting). Indeed, writting style sometimes seems to fall backwards with recent releases.

Also, it's subjectively, but commonly held to be true that most of the content created when AI can execute well lack taste, unless the human creating with AI has it. The term "slop" exists and in legible for a reason.

I mean, maybe this is just a question of different people speaking about different time horizons.

But it seems perfectly sensible for someone to say that there is a lot of automation that can happen in the near term, and that they see no reasonable evidence that AI can do interesting or even coherent work this space.

I don't know that shallow accelerationist boosterism (or panicked doomer despair ) that handwaves away the distinctive variations AI performnce is that enlightening.

The way I see it, it’s all about how you use it. Let people have tools; their output stands on its own legs.

As pro software, the UI has a learning curve. Many YouTube videos are out of date, or unnecessarily drawn out.

Additionally, as a hobbyist, project settings, color spaces, clipping, etc can be tricky to miss.

I’ve already been using Claude to (1) help me with UI, (2) QA my delivery, and (3) occasional fixes for audio repair; waveform sync (can’t afford time codes).

I make all creative decisions - including most menial ones - myself, I wouldn’t have it any other way; I enjoy the act of creating. I welcome this integration.

I’m sure some users are gonna agentically grade or whatnot. I don’t like it, but I’m also don’t believe in telling other people what they should do with their own work and creative process.

Because there has to be demand for it. There are hundreds of tasks that could be automated right now, and people still choose to do them by hand. Just because something is technically replaceable doesn’t mean it will be replaced and editing is a deeply personal part of the process for almost every director. It will change, but for serious work, it will never be a sure shot job handled entirely by AI. Sure, for low effort marketing videos you’d outsource to Fiverr, it’ll happen sooner or later but that’s not what we’re talking about here.

> But this whole line "but it won't replace the real experience and taste required for larger projects" is not how I see it.

Why not? Personally (although I'm not saying others have to see it this way too), I'd wait until I can see what AI can replace, before believing so, rather than the opposite.

So far, AI isn't able to "set direction" at all in any sensible way, especially for a creative project like TV or movies, the results are just too bad (today). Although it's excellent at "following directions", which is why these new APIs are great and exciting, it lets people who are good at setting direction to leverage these tools to have the direction concretely followed by something, without having to create a group of collaborators.

Until we've seen proven that AI can indeed "set directions" ("have taste"), I think it's fair to assume they're unable to do so in a way that it produces something that humans generally enjoy. And very far from the kind of masterpieces a human can create.

I understand what you're saying and I agree, ostensibly, but on the caveat that it's hinged on the "we don't know what we don't know" aspect of reality. If we restrict our domain to what's forseeable, it's still apparently the case. Astra's demos have me scratching my head, diminishing returns are kicking in harder and faster than anticipated. It seems we've entered a phase to where frontier companies have completely abandoned improving the substrate because they can't justify it. It seems like it's all just harnesses and reasoning orchestration at this point. Hell, everybody but Google seems to have given up even on projection layer improvements (and Gemini is unusably bad in other ways).

Watching this next generation closely, because there are just a shitload of red flags starting to crop up that don't seem to be going away.

Yes people really need to take the AI’s feelings into account before they say things like that.

Taste is subjective - that's why it can't be replaced.

I would gladly enjoy to collaborate with an agent if I can learn more about video editing. I edit my video only for passion and to learn something new though, so I guess that professionals may see this in a different way. But I don’t think that it’s much different from how Claude code revolutionised the software scene.

I would not be surprised if engineers who laid the groundwork for this back then would be happy with this workflow.

I am reminded of an interview I saw about the Stranger Things intro (before mainstream audiences got weird about the show). The producers wanted a practical effects intro, and consulted with people who actually made 1980s intros the show was trying to replicate.

The old guard said the show’s producers were crazy to do it with practical effects and should just use CGI since this was what they would have done, if they had the technology.

Edit: oh god, there is a nearly identical HN post from 2024 from me that I do not remember making. I swear I am not an LLM. I am also hoping that I am not an old man telling his stories over and over either: https://news.ycombinator.com/item?id=40453073

If we can get the environmental impact of this stuff ameliorated (a big if), this is the real benefit of this kind of AI assistance imo- it transforms everyone into a director and frees them from being a surrogate for a director.

See also: 3D modeling. Autodesk Fusion's interface is truly awful and people put years into learning it to do complex things. The MCP server functionality they added, though still in its infancy with Claude and what it can do, vastly lowers the barrier to entry for people just looking to get into making basic things. They can focus on the actual creative process rather than fumbling with the awkward tooling.

>Not every application of AI is "do the creativity for me."

However it can lead to AI "doing the creative for them" and them out of a job in the near future.

I'm not complaining either, I wouldn't call myself a video editor, but I do end up spending some of my time in Resolve, and anything to remove the dreadful monotonous tasks of editing interviews is welcome. Although Resolve already makes it pretty easy, with the "edit by moving the transcript around" feature which is amazing. Anyway, I'm happy about being able to sick Codex at Resolve now :)

That's why there are no assistant editor jobs anymore.

Pretty soon there will be no need for recording and editing .

IIRC House MD was one of the first shows to use digital (5D mkII).

Soon there will be AI generated shows. Perhaps entirely personalized.

Youtube, generate me a TNG episode.

Resolve has had de-aging filters, and TTS voice synthesis for awhile.

If the chat-bot/spy-ware aren't integrated, than I don't see a problem with LLM plug-in opt-in support. =3

So let them? A product that focuses on chasing the competition is effectively useless, because the competition will always be ahead.

I can imagine the conversation with Studio lawyers about contaminating IP with uncopyrightable generative content liabilities. =3

Is it surprising that they’d rather make features that cater to a large portion of their user base instead of a tiny fraction of it?

did they remove it based on policy? seems.. less convenient.

I think there's a really interesting tension between what you're saying and what the parent post is saying, and it's something that's been discussed for a long time but LLMs are throwing a new wrench into it: a given user probably only uses 5-10% of the features of a piece of software, but across the whole body of users you get close to full coverage.

I use DaVinci Resolve for editing GoPro, Canon DSLR, iPhone, and Drone FPV footage. I import clips, use the razor to cut them and rearrange them, occasionally crop and zoom, and then encode them down to h264 or whatever. There are a bazillion features that I sort of know exist but haven't ever had any need to use. Could I make a tool using an LLM that covers my DaVinci workflow? Almost certainly, and I don't think it would take anywhere near 200 prompts. I've made significantly more complex bespoke data analysis pipelines for drone flight telemetry in like... a dozen prompts. But I have no illusions that what I would make as my "DaVinci Clone" would have anywhere near the functionality of DaVinci, just the tiny piece of that functionality that I use.