This post is adapted from my conversation with Andy Coon on the Talent Freaks podcast. Andy is a filmmaker and photographer based in the Bay Area. We covered a lot of ground, from building Vindexr to filmmaking philosophy to the right way to think about gear. The full episode is worth watching.
Andy had tested a beta build of Vindexr for about a week before we recorded. He ran some of his podcast footage through it, typed in a few searches, and came away impressed. He also said something I hear pretty often: "I can see exactly where this would have saved me, at a previous job where we were always hunting for B-roll from three or four years ago. You remember the shot, but finding it is a whole different problem."
That's the thing about footage management. The problem only gets visible once you have enough material to drown in.
The Two Friction Points
I've been in video production for almost twenty years. Corporate, agency, client-side, back and forth. And editing always ends up being the bottleneck. Shooting is fast. Editing is the four-minutes-per-minute-of-footage part that nobody warns you about.
When I started thinking about what actually made editing slow, it came down to two things.
The first is finding the clip. If it's recent footage from a known project, fine. But if you're going back into an archive, you're digging through drives and folders and trying to remember the shoot date. And even once you find the right folder, video files are named with camera prefixes and strings of numbers. That tells you nothing. So you're scrubbing through takes to find what you need, and that's before you've done any actual editing.
The second problem is extracting from the clip what you want. Even after you've located the right interview or the right scene, it's a linear recording. You have to put it on a timeline, play it back, listen again, find the usable section, mark in and out points, cut it out. It's all manual. It all lives in your short-term memory while you're doing it.
Vindexr is built to remove both of those friction points.
What the Demo Showed
During the episode, I walked Andy through a real project I was working on: a set of interviews with engineers at my day job, talking about lean manufacturing principles. Not the most cinematic subject matter, but a good test case. The goal was to cut several long interviews down into a three-minute internal video.
Once the footage is imported, Vindexr auto-transcribes the audio and runs machine learning models on the visuals to tag what's in each frame. Subjects, objects, environments. That all becomes searchable metadata. You can click through a transcript that syncs to the video player, or export it as a text file.
But what I showed Andy that surprised him was the collection system. You can set in and out points while you're listening, hit a key to clip that segment to a collection, and build what amounts to an assembly edit without ever opening your editing software. You can preview the collection, reorder the clips, and pass that structured selection over to your NLE when you're ready.
Andy stopped me when I explained that part: "Wait, you're rough cutting inside the tool?" That reaction is pretty common. It's one of those things that sounds like a small feature until you're actually using it.
The other half of the demo was the AI query interface. I can take all the metadata Vindexr has gathered from a folder of footage and pass it to the Claude API. Not the video itself, just the metadata. So there's no expensive upload, no waiting for cloud processing. I already have the transcripts and visual tags locally. Passing those to Claude and asking it to find the best sound bites and arrange them in story structure took seconds.
I showed a prompt in real time: "Find all the best sound bites and arrange them in good story structure." The results came back as timestamped clips pulled from across multiple interviews, organized into a narrative arc.
Four Days to Four Hours
Andy had seen something I'd posted on LinkedIn. Six interviews, material that would normally take four days to cut down to a five-minute piece, done in four hours.
I was honest about it on the podcast: I might have been exaggerating a little. But not by much. And the real explanation isn't just speed. Part of what made those four days feel like four days, after almost twenty years, is motivation. Going through raw interview footage and manually hunting for the compelling moments, again, for the hundredth time, is genuinely hard to get motivated for. When the tool can surface the structure of the material for you and you can walk in already knowing where the story lives, the edit itself gets interesting again.
That's the framing I keep coming back to. If we can use AI to handle the parts that feel like a spreadsheet so that humans can spend more time on the parts that feel like creativity, that's the right use of the technology. Vindexr isn't trying to make editorial decisions. It's trying to get you to the point where you can.
The Local Processing Piece
Andy brought up something I hadn't expected to spend time on: security. He works with corporate clients where plugging in a USB drive requires IT clearance. Sharing files is paperwork. Cloud uploads with sensitive material can be a compliance nightmare.
I work at Idaho National Laboratory, which is a Department of Energy facility. The security environment there is even more layered. We've only recently gotten access to AI tools at an enterprise level, and the path to get there was complicated. What you can upload, what you can share with a third-party tool, all of that is heavily scrutinized.
Local processing isn't just a technical architecture decision for Vindexr. For a lot of the people who need this kind of tool most, it's the only way the tool is usable at all. The footage never leaves the machine.
Obsess Less About Gear, More About Light
The conversation shifted later in the episode toward filmmaking philosophy, and I want to include some of that here because it connects to how I think about Vindexr too.
Andy asked what I'd tell younger DPs to obsess about less and what to obsess about more. The answer for less is obvious: gear. We all fall into that trap. If I were just shooting on a RED, then I'd have arrived. That's not how it works. If you haven't grown as a filmmaker, the RED camera just sits there.
What to obsess about more is light. Learn to see it. Learn to shape it. That's more transferable than any piece of glass or any camera body.
The reason I think this connects to the tool is the same underlying idea. The gear is not the creative work. The search problem is not the creative work. The logging is not the creative work. What's left after you've removed those frictions is the actual thing worth doing.
Andy made a point toward the end of the episode that I thought was well-put: when you're working in corporate video, you're not just an editor. You're the producer, the shooter, the director, sometimes the set designer. Turnarounds are fast. You don't have time to go as deep as you want on any one piece. What Vindexr gives back is the space to do some of that. The color work. The sound design. The parts that make something feel considered.
The full episode is on YouTube here. Andy and I also got into the Area 51 documentary we shot in 2019, the self-worth trap that creative fields set for you, and a fairly entertaining story about a RED camera hitting concrete on a shoot. Worth the watch.
If Vindexr sounds like something that would fit your workflow, reach out. I'm happy to let you try it.