I Tried Fully Automating My Video Editing With AI (RAW RESULTS)

Transcript

[00:00] Oh my god, Opus 5.5 is incredible at
[00:02] making launch videos. Look at this one
[00:03] shot. How far can I go trying this sort
[00:06] of AI editing compared to our human
[00:08] video editor who is very good. If this
[00:09] works, I will be absolutely amazed. That
[00:12] is really good. Oh no. Oh no. I’m
[00:15] worried for video editors. We’ve got
[00:17] amazing video editors on my team.
[00:20] So, every time a new AI model comes out,
[00:22] my Twitter feed fills up with people
[00:24] saying that this one can now edit
[00:26] videos. And I’ve been seeing that with
[00:27] basically every model release for the
[00:29] last few months. And so I was a little
[00:30] bit skeptical. Now, for context, I
[00:32] edited my own videos for about three
[00:33] years back when I was working as a
[00:35] doctor for somewhere between 10 and 40
[00:37] hours a week. And so I kind of know what
[00:38] a good edit feels like, but I’d never
[00:40] actually sat down and properly tried
[00:42] editing with AI. But then a public
[00:43] holiday came around and the team mostly
[00:45] had a day off and we needed a launch
[00:46] video for Creator Grade, one of our
[00:48] platforms. Path, one of our editors, was
[00:49] already doing it the normal way. And so
[00:51] I decided to challenge myself for the
[00:53] day. Could I, as someone who’s never
[00:54] tried AI video editing, learn how to use
[00:56] AI for video editing and make a video
[00:58] that comes as close as possible to a
[01:00] world-class human editor? And to be
[01:02] honest, if AI could get me like 80% of
[01:04] the way there, I would be delighted.
[01:05] This video is a little bit experimental.
[01:07] Usually, I tell you about stuff that I
[01:09] already know. But in this video, I’m
[01:10] taking something that I don’t have any
[01:12] idea about. And so, if you want, you can
[01:13] watch the process of how I’m sort of
[01:15] learning it and figuring it out. And
[01:17] hopefully there’ll be something that you
[01:19] can take away from this to apply to your
[01:20] own workflows for learning things or for
[01:22] applying AI. This is going to be a super
[01:23] long video, so there’ll be timestamps
[01:25] down below. Feel free to skip around if
[01:26] you like. But now before touching
[01:27] anything, I tried to work out what video
[01:29] editing actually is because that’s how
[01:31] we’re going to figure out how to break
[01:33] it down using AI. What is video editing
[01:35] at its core? Okay, so if we think of
[01:37] when you’ve already got some footage
[01:38] compared to when you don’t. So when
[01:39] you’ve already got some footage, what is
[01:40] video editing? Well, firstly you import
[01:42] the footage and then you need to
[01:44] basically cut out all the mistakes to
[01:46] get a clean a cut. Then once you’ve got
[01:48] a clean a cut, you then add in B-roll.
[01:51] It would be like the screen recordings
[01:53] and the screenshots and stuff. Then the
[01:55] hard part, the bit that takes ages is
[01:57] animating the B-roll so that you’ve got
[01:59] like zoom in and zoom out instead of
[02:01] just showing a screenshot which would be
[02:03] boring. These launch videos often go in
[02:04] and out and then there’s motions and
[02:06] transitions and fadeins and movements
[02:07] and swipes and pans and all those
[02:08] things. So that takes a human video
[02:10] editor ages to do. Once you’ve done
[02:12] that, then it’s like background music
[02:14] and sound design. So the background
[02:16] music, you know, we’ve got a particular
[02:17] vibe, but then sound effects like risers
[02:20] and swipes and swishes and all that kind
[02:21] of stuff. That is ultimately what video
[02:23] editing comes down to. And then
[02:24] obviously the taste factor is telling a
[02:27] good story and deciding what to cut out,
[02:29] deciding what to keep in. I feel like
[02:31] once we break it down into all those
[02:32] things these days, surely the AI should
[02:34] be pretty good at all that stuff. To be
[02:35] honest, I would be very amazed if it
[02:37] manages to just even get the a cut right
[02:40] immediately, right? Cuz when I, as a
[02:41] human, I’m editing a video, if there is
[02:43] a pause or there is a retake, you have
[02:45] to know where to cut and splice the
[02:47] footage together so that it doesn’t
[02:49] sound weird. You know, we had a whole
[02:50] lesson in my course where I taught
[02:51] people how to use Final Cut Pro. There
[02:53] was a lot of time spent on just like
[02:55] nailing the A cut cuz if you can nail
[02:56] the A cut, then everything becomes
[02:58] smoother further down the line. But if
[02:59] the A cut sucks, then everything else
[03:01] will kind of feel a bit disjointed. The
[03:03] stuff I’ve seen on Twitter is like super
[03:05] super super polished. They’re like, “Oh
[03:06] my god, Opus 5.5 is incredible at making
[03:08] launch videos. Look at this one shot.”
[03:09] And I’m like, “There’s no way that was
[03:11] actually one shot. Surely you had to go
[03:12] back and forth quite a lot and do a lot
[03:14] of tinkering to get that particular
[03:15] result.” So, I’m going to be very
[03:16] interested to see what does the oneshot
[03:18] look like where I’ve just I I’m not
[03:20] really driving it. I’m just like giving
[03:21] it copied and pasted stuff from Grock
[03:23] into cloud code and then claw code into
[03:24] claw code just like raw dog and just see
[03:26] what it does. And then when I’ve dabbled
[03:28] with AI stuff just for anything in the
[03:29] past, getting it to like 80% good enough
[03:31] happens with one shot. But then
[03:33] polishing that final 20% takes freaking
[03:35] ages and the bulk of the time is spent
[03:37] in polishing it so it feels good and
[03:41] looks good and has, you know, taste
[03:43] applied to it rather than just the
[03:45] oneshot version.
[03:48] All right, so phase one of this project
[03:49] is I want to start by getting the lay of
[03:52] the land. What options do we even have
[03:54] for AI video editing? So, I’m just going
[03:56] to use drug Grock cuz I don’t need a VPN
[03:58] for it. I’m trying to figure out how to
[04:01] do AI video editing. I’ve got some
[04:03] footage I need to turn into a launch
[04:07] video for Creator Grid platform that
[04:10] helps pitches learn grow and monetize. I
[04:13] want to see if I can use AI video
[04:16] editing to make a alternative edit.
[04:21] Ideally, the whole thing, but also we
[04:24] could do a 30 to 60sec faceless launch
[04:28] trailer type thing using motion graphics
[04:31] and stuff. Or we could do it like
[04:34] talking head with screen shots like a
[04:38] YouTube video, some combo of all of the
[04:41] above. Help me figure this out. Ah,
[04:44] three different pipelines. So, one
[04:46] option is the Da Vinci thing with an
[04:49] API. Okay. Descript for the cut and then
[04:51] resolve for grade graphics and finish.
[04:53] Okay, option three. This is the remotion
[04:55] hyperframes lane faceless brand launch
[04:58] trailers. Cool. Not entirely sure what
[04:59] any of this means just yet, but I think
[05:02] we have a reasonable lay of land. All
[05:04] right, so quick translation because
[05:05] there was a lot of jargon over there.
[05:07] There are basically two ways AI can make
[05:09] you a video right now. The first one is
[05:11] to edit the footage that you’ve actually
[05:13] filmed. For that, I used Da Vinci
[05:14] Resolve, which is a proper professional
[05:16] video editor that recently started
[05:18] letting AI control it directly. But then
[05:19] the second is to not edit footage at
[05:21] all. You instead get the AI to write the
[05:23] video as code and it generates every
[05:25] single frame. That’s what tools called
[05:27] hyperframes and remotion do and it’s how
[05:29] a lot of those launch videos that you
[05:30] might have seen on YouTube and Twitter
[05:32] are being made. I didn’t know which
[05:33] approach would work and so I decided to
[05:35] try both of them at the same time. Oh,
[05:37] what a vibe. Right as of literally right
[05:39] now, Path, one of our human video
[05:40] editors, has actually just sent me the
[05:42] first draft of this video that he was
[05:45] manually editing. So, I don’t want to
[05:47] watch it right now, but that could be
[05:49] interesting to compare. How far can I go
[05:52] with trying this sort of AI editing
[05:54] stuff for the first time compared to our
[05:56] human video editor who is very good.
[05:59] Okay, awesome. I’ve got a roll filmed
[06:02] already. How do we get started? Okay,
[06:05] firstly, Hen Hyperframes core skills
[06:08] experiment B A Cut Resolve MCP needs Da
[06:10] Vinci Resolve Studio 21.1 295. All
[06:13] right, guess I’m paying 300 and it is
[06:23] an 8.5 GB file. So, it is downloading.
[06:26] All right, now I need to Claude plug-in
[06:28] marketplace. Hen hyperframes. Install
[06:30] Hen Hyperframes. Let’s open up Claude
[06:32] code. Uh, so what I want to do is let’s
[06:35] go local. Let’s make a new folder. uh
[06:39] creator grid launch video or something
[06:43] and let’s open 5.5. Okay, I’m trying to
[06:47] use hyper frames to make a launch video
[06:50] for creator grid. I’ve already got some
[06:53] a roll, but we don’t necessarily need to
[06:56] use it. Here are some pre notes from
[07:00] Grock. Let’s see what Claude does. Okay,
[07:03] we are installing hyperframe skills into
[07:05] the project. All right, so I’ve already
[07:07] got that footage. Should I just chuck it
[07:09] into the project folder? Oh, and by the
[07:11] way, I’ve added in the existing video
[07:14] into the project folder. FYI. Ah, okay.
[07:16] Claude code is asking ffmpeg isn’t
[07:18] installed. Yes, by all means, please do.
[07:21] Yes, install ffmpeg. Happy for you to
[07:24] install it globally on this machine.
[07:26] Pablo, our co-founder, has put together
[07:29] this claude artifact with product views
[07:33] that we can use. I want to see what a
[07:35] pure faceless just music
[07:38] no uh voice video looks like, but also I
[07:42] want to see what it looks like with just
[07:45] my voice as a voice over and also what
[07:48] it looks like with both my voice and
[07:50] face because I think my face and voice
[07:54] carries trust to some degree. Let’s do a
[07:56] pipeline test first. It’s doing stuff.
[08:00] Installing FFmpeg and Whisper CPP. I
[08:03] know what both of those things mean, so
[08:05] I don’t need to ask. But if I didn’t, I
[08:07] would totally ask. The screenshots are
[08:08] good. Real UI at 4K. Setting up
[08:09] Hyperframes project and running the
[08:11] 10-second pipeline test. Great. It’s
[08:13] just doing stuff. Oh, quick note from
[08:15] the sponsor of this video, which is
[08:16] Whisper Flow. Now, on a good day, I can
[08:18] type at about 156 words per minute. But
[08:20] even with that insane typing speed, I
[08:22] can still speak faster than I can type.
[08:23] And so, Whisper Flow is a wonderful tool
[08:25] for turning your voice into text in
[08:26] pretty much any app you might be using.
[08:28] So, for example, I’m going to try this
[08:29] right now in Slack. Hey, Angus, we’ve
[08:31] got three things for Austin. Uh firstly,
[08:32] can we book the flights by Friday? Hang
[08:33] on. Uh Thursday, actually. And uh make
[08:36] sure everyone’s landing the evening
[08:37] before. Second, um can you please send
[08:39] everyone the agenda with the planning
[08:40] sessions in the morning and the
[08:41] afternoons left free? And then thirdly,
[08:43] can we sort out a group dinner, please,
[08:44] for the first night? Ideally, somewhere
[08:45] that we can all fit in one table. Cool.
[08:47] Thank you so much. Okay, so that took me
[08:49] under 30 seconds. And if you look at
[08:50] what’s come out, it’s turned my spiel
[08:52] into a very nice numbered list that
[08:53] makes it very easy for the person to see
[08:55] what I’m talking about. I made a mistake
[08:56] on saying Friday, no wait Thursday, and
[08:58] it took that into account. Now, people
[08:59] always say that, “Yeah, but my phone’s
[09:00] got built-in dictation.” To which I
[09:02] would say, “Have you seen how bad the
[09:03] built-in dictation on your phone is?
[09:04] Let’s just try it out.” Uh, “Hey, Tam.
[09:06] Uh, are you around for a call on
[09:07] Saturday?” No, wait, uh, Sunday. And,
[09:09] uh, can you send me the name of that
[09:10] book that you mentioned? So, yeah, it
[09:12] doesn’t know how to spell my brother’s
[09:13] name. It doesn’t clock the mistake. And
[09:15] now, if we try exactly the same thing in
[09:16] whisper flow. Hey Tam, uh, are you
[09:18] around for a call on Saturday? Um, no,
[09:19] wait, Sunday. And, uh, can you send me
[09:21] the name of that book you mentioned?
[09:22] Great. And it even knows that this is
[09:23] exactly how I spell my brother’s
[09:25] nickname. His name is Tamour, and I call
[09:26] him Tame. And it knows exactly how to
[09:27] spell it. and it’s automatically taken
[09:29] the error into account and removed it
[09:30] from the text, which is awesome. In my
[09:32] case, I’ve currently written 167,630
[09:36] words on Whisper Flow so far, which is a
[09:38] fun little stat to see. Anyway, Whisper
[09:40] Flow is free to use, and if you want to
[09:41] use it absolutely loads, then you can
[09:42] get Whisper Flow Pro. That gives you
[09:44] unlimited words, and you can get a month
[09:45] of Whisper Flow Pro completely free by
[09:47] checking out the link in the video
[09:48] description. And when they ask you how
[09:49] you heard about them, please do write
[09:51] YouTube and put my name in the how did
[09:53] you hear about us box because then
[09:54] they’re more likely to sponsor more of
[09:55] our videos, which is awesome. So, thank
[09:56] you Whisper Flow for sponsoring this
[09:57] video. and let’s get back to it. Okay,
[09:59] awesome. Have we downloaded Da Vinci
[10:01] Resolve yet? Yes, we have. Let’s install
[10:03] this. I’ve literally never worked with
[10:05] Da Vinci Resolve, but how hard can it
[10:08] be? Okay, classic. You look at a video
[10:11] editor and it looks intimidating AF.
[10:13] What do I do here? Help. AI. Set up AI
[10:17] assistance. All right, let’s do that. Da
[10:19] Vinci has been added to these. [music]
[10:20] Please res Oh, wow. Okay, cool. It’s
[10:22] already done. So, claw desktop extension
[10:24] extension installer opened. Da Vinci
[10:25] Resolve Studio extension. Okay, then
[10:28] install. Let’s go. Oh, that was quick.
[10:31] Uh, let’s see what’s Claude Code saying
[10:33] now. Still doing stuff. It’s using some
[10:36] tokens. 3% of session limit. All right,
[10:37] we’ve got plenty of plenty of headroom
[10:39] here. So, what the hell do I do next on
[10:41] Da Vinci? I mean, in theory, it should
[10:43] be connected to Claude code, right? So,
[10:45] I could just Do you see the Dainci
[10:49] resolve connection MCP thing? Let’s see
[10:52] what happens. Yes. Wow, we’re about half
[10:54] an hour in. So far, we’ve managed to buy
[10:57] and download Da Vinci Resolve Studio and
[10:59] Claude Code is cooking away on trying to
[11:02] do the hyperframes method. By the way,
[11:05] can you write me a brief I can put in
[11:09] another Claude code chat within this
[11:11] project to test out Dainci Resolve MCP
[11:15] for building this video as an
[11:18] alternative method. Nice. All right.
[11:21] Well, claw code is basically telling me
[11:23] what to copy and paste into claw code.
[11:25] Boom. I love a good old copy paste from
[11:30] AI into itself. Oh, wow. Okay, cool.
[11:35] Storyboards are written. No animation
[11:36] code yet and your bands gradients,
[11:38] glow-ins, fade in, blah blah blah. Film
[11:40] a music and text 30 seconds. This is its
[11:42] [music] own edit with no voice. The type
[11:44] and the product shots have to tell the
[11:45] whole story. Okay. Yep. Film B, your
[11:47] voice 45 seconds with a 7second extended
[11:49] cut. Build it once, render it twice.
[11:51] voice only and then uh with your face
[11:53] added film a silent has three options.
[11:55] Okay, let’s look at that. Oh, okay. It’s
[12:00] fun. All right, let’s just ask these.
[12:02] Um, one, demo data. Two, yes, we can
[12:05] show client logos. Three, that’s okay.
[12:09] Four, is there a logo file? Um, get logo
[12:12] from creator grid.io. Yes, static. No,
[12:15] animation. Still render to prove the
[12:18] setup works. Please pick an A option and
[12:20] the B length. Okay, so B length 45
[12:23] seconds feels good. And A, you do two
[12:27] things, we do everything else. All
[12:28] right, creative marketing without the
[12:29] chaos. Nice. I quite like that. That’s
[12:31] really good. A1 option I like. Let’s go.
[12:35] Hey, this is so good. Okay, now let’s
[12:36] see what Resolve’s doing. Creating work
[12:39] folders. Find whisper models. Probe
[12:41] source view live output. Resolve
[12:43] scripting API. All right, it’s just
[12:44] doing stuff. I guess maybe I won’t even
[12:46] need to use. It’s not visibly doing
[12:48] anything, but [music] I guess it’s doing
[12:49] stuff in clawed code with the Da Vinci
[12:51] Resolve connector. If this works, I will
[12:53] be absolutely amazed. There’s no way
[12:55] it’s going to work first time. There’s
[12:57] just no way. But this is quite cool.
[12:59] It’s such a vibe, isn’t it? Oh, open and
[13:01] split view. Oh, wow. [gasps] This is
[13:04] really cool. So, at this point, we’ve
[13:06] got stuff happening basically side by
[13:08] side. So, on this side of the screen,
[13:09] we’ve got the Da Vinci Resolve method of
[13:13] trying to do AI video editing. On this
[13:15] side of the screen, we have the co
[13:16] claude code hyperframes method of doing
[13:19] AI video editing. And I’m just sort of
[13:22] sitting back and waiting for it to like
[13:24] do stuff. Uh, which feels pretty good.
[13:27] Yeah, this IMAX fans are starting to
[13:30] starting to get pretty Yeah, this is
[13:32] hot. Everything’s hot. If this really
[13:34] works, I might just upgrade to an M5 Max
[13:35] Studio. Oh, wow. Look at that. That is
[13:37] actually not bad. Creator marketing
[13:39] without the chaos. That actually looks
[13:40] pretty freaking solid. That looks pretty
[13:42] good. I mean, as anyone can generally
[13:43] generate a still these days, but let’s,
[13:44] you know, remains to be seen how good
[13:46] this is for video editing. Let’s see
[13:47] what happens next. All right, both story
[13:50] boards are sketched. No animation code
[13:51] has been written yet. Okay, cool. Let’s
[13:53] look at these story boards. This
[13:54] actually remarkably good. Creative
[13:56] marketing without the chaos. I don’t
[13:58] like this font, but I think this is our
[13:59] font. Is that our font? Okay, I might
[14:01] have to redesign the creative grid
[14:02] branding if we do use that font. I’ll
[14:04] get to the style later. Pick your
[14:05] creators. Prove the content. We handle
[14:07] the in sourcing, negotiating, time
[14:09] briefing, reviewing, invoicing, paying,
[14:10] tracking, everything else. Nice. Okay.
[14:13] So far, I’m I am quite impressed by this
[14:15] storyboard. I wonder if I should use
[14:17] storyboards more when doing stuff.
[14:18] What’s happening on Da Vinci side? Okay,
[14:20] it’s still doing stuff. Okay, so Da
[14:22] Vinci has been cooking for 20 minutes
[14:23] and this just seems to be running into a
[14:25] bunch of issues. I probably should have
[14:27] done this one at a time, but you know
[14:29] what? Let’s just see. Let’s just see
[14:30] what it’s going to do. One shot. You
[14:32] know, one of the things with working
[14:33] with AI is figuring out what to do while
[14:35] it’s while it’s thinking. I could try
[14:37] and do a third thing. Why not? By the
[14:39] way, I want to try using reotion as a
[14:42] third experiment running concurrently
[14:45] with hyperframes and resolve. Write a
[14:48] brief that I can give to a new clawed
[14:51] code instance within this project for
[14:54] this. Boom. So, we’ve got hyperframes
[14:56] number one. We’ve got resolve is method
[14:58] number two. And we’ve got remotion which
[15:01] is method number three. So we now have
[15:03] hyperframes over here, resolve in the
[15:05] middle, and then reotion over here.
[15:13] Okay, all three drafts are rendered.
[15:15] [gasps] We have MP4 files. All right,
[15:18] let’s see what does this look like and
[15:20] is it any good?
[15:36] H. Okay. Well, that’s pretty bad as a
[15:39] faceless version. Let’s try Draft V2
[15:41] voice. No face. Let’s see what happens.
[15:43] If you’re a brand trying to grow with
[15:45] creator marketing, you probably know
[15:47] that it is a total nightmare. If that
[15:49] resonates, I would love to tell you
[15:50] about Creator Grid, where we will
[15:52] literally run your creator campaigns for
[15:54] you end to end. You don’t pay anything
[15:56] upfront. There is no retainer. There’s
[15:58] no subscription. And you only pay for
[16:00] results. Entrepreneur and creator. If we
[16:02] haven’t met, hello. I’m What the hell?
[16:05] Ali, I’m a doctor turned entrepreneur
[16:07] and creator with about 8 million.
[16:09] Creator Grit is our attempt to basically
[16:10] give you the level of service. Oh no. Oh
[16:13] no. It didn’t. It screwed up the A cut.
[16:16] Just the basic A cut. It just doesn’t
[16:18] get it. I knew it. I knew it. The
[16:20] pauses.
[16:21] The pauses. The pauses are bad. It’s not
[16:24] It’s not doing the A cut properly. And
[16:25] now, what if we have my face in it as
[16:26] well? What does that look like? If
[16:27] you’re a brand trying to grow with
[16:29] creator marketing, you probably know
[16:31] that it is a total nightbeck. If that
[16:33] resonates, I would love to tell you
[16:34] about creator grid where we will
[16:36] literally run your creative campaigns
[16:39] for you end to end. You don’t pay
[16:42] anything up front. There is no retainer.
[16:43] There’s no subscription. And you only
[16:45] pay for results. entrepreneur and
[16:47] creator if we haven’t met. Hello, my
[16:49] name is
[16:51] this this is bad. But this would have
[16:55] been magical a few years ago that like
[16:56] wo you could just do that with like just
[16:59] getting a robot to operate Da Vinci
[17:00] Resolve. But I had way higher hopes of
[17:03] the V1. Okay, hopefully Hyperframes does
[17:05] a better job with this. So, I think what
[17:07] I may have to do is I might have to
[17:10] actually manually decide what to keep in
[17:12] in terms of what I’ve said or maybe even
[17:14] record a separate video that’s like
[17:15] explicitly word for word scripted 45
[17:18] seconds long cuz this one was more like
[17:20] 5 minutes long and it’s just tried to
[17:21] figure out which bits to Hm. This is
[17:24] frustrating. I do not like this. Let’s
[17:26] grab something from Epidemic Sound. I
[17:29] want to find a track that people would
[17:31] recognize.
[17:39] Oh, beautiful. Right. Okay. So, we’re
[17:41] using Valdi’s spring as the background
[17:43] music for the hyperframes one. Okay.
[17:47] Still we’ve got hyperframes doing stuff.
[17:48] We’ve got Resolve doing stuff and now
[17:50] we’ve got Remotion that’s been doing
[17:51] stuff for the last 23 minutes. Also, one
[17:53] thing that I didn’t really do is I could
[17:55] have like I’ve been seeing stuff on
[17:56] Twitter recently about that where people
[17:58] have come up with these like elaborate
[17:59] prompts for hyperframes. So, maybe I
[18:02] could have just copy pasted one of those
[18:03] just to see what that would do. But
[18:04] we’re not going to get ahead of
[18:05] ourselves. We’re just going to try and
[18:07] focus on the the simple stuff. We have a
[18:10] silent draft of the hyperframes video.
[18:13] Shall we see what this looks like?
[18:43] >> [laughter]
[18:45] >> Okay. It It’s It’s not bad. The
[18:48] classical music makes it kind of funny.
[18:50] Um, [laughter]
[18:51] it’s a little bit janky. It’s the
[18:53] equivalent of if I were like 15 years
[18:56] old and learning video editing for the
[18:58] first time and deciding to use Vivaldi’s
[19:02] spring as the background music. It’s
[19:04] impressive as a V1, but of course we
[19:07] need refinements to make it actually
[19:08] good. Okay, I’m going to go for lunch.
[19:10] We’re going to come back and my
[19:12] prediction is that I will be
[19:13] disappointed with the oneshot results of
[19:15] this and then we will done. We will
[19:17] start the actual work of doing anything
[19:19] with AI which is polishing
[19:23] All righty, ladies and gentlemen. We are
[19:26] back after our lunch break and it is
[19:28] time to see what is happening. So, we’re
[19:31] going to start with the hyperframes one.
[19:35] So, we saw the silent one a little bit
[19:37] earlier. Let’s try voice and no face. If
[19:40] you’re a brand trying to grow with
[19:41] creator marketing, you probably know
[19:43] that it is a total nightbeck. I would
[19:45] love to tell you about creator grid
[19:47] where we will literally run your creator
[19:49] campaigns for you end to end. All you
[19:51] have to do is basically two things. So
[19:52] firstly, you decide which creators you
[19:54] want to work with. So we’ll source them
[19:55] for you. You just sort of approve or
[19:56] reject them. We will give you a bunch of
[19:58] content and you just decide to approve
[20:00] it or not. Everything else is handled by
[20:02] us. So the sourcing of creators,
[20:03] negotiating, reviewing the content
[20:04] initially, and then invoicing and paying
[20:06] the creators, tracking all the views,
[20:08] blinds, all of that annoying stuff we
[20:10] take off of your plate. Oh wow. Okay,
[20:12] that’s like remarkably impressive.
[20:14] That’s way better than the uh Da Vinci
[20:17] Resolve one. That’s really well done.
[20:19] Uh, let’s see what this looks like with
[20:20] my face as well.
[20:21] >> If you’re a brand trying to grow with
[20:22] creator marketing, you probably know
[20:24] that it is a total nightmare. I would
[20:26] love to tell you about creator grid
[20:27] where we will literally run your creator
[20:29] campaigns for you end to end. All you
[20:31] have to do is basically two things. So
[20:33] firstly, you decide which creators you
[20:34] want to work with. So we’ll source them
[20:36] for you and you just sort of approve or
[20:37] reject them. We will give you a bunch of
[20:39] content and you just decide to approve
[20:41] it or not. Everything else is handled by
[20:42] us. So, the sourcing of creators,
[20:44] negotiating, reviewing the content
[20:45] initially, and then invoicing and paying
[20:47] the creators, tracking all the views,
[20:49] blinds, all of that annoying stuff we
[20:51] take off of your plate.
[20:52] >> Wow. Okay, that is very good. I mean,
[20:56] obviously, it could do with a lot of
[20:57] work, but that is remarkably impressive.
[21:00] I did not think it would be that good.
[21:03] Let’s see how Remotion does it. Wa!
[21:06] Remotion has literally given me a video
[21:09] editor interface. What the hell? I
[21:11] didn’t realize I could do that. That’s
[21:12] cool. Okay, let’s see what this version
[21:14] looks like.
[21:15] >> If you’re a brand trying to grow with
[21:16] creator marketing, you probably know
[21:18] that it is a total nightbeck. If that
[21:21] resonates, I would love to tell Hang on,
[21:22] let’s try the voice face one. If you’re
[21:24] a brand trying to grow with creator
[21:25] marketing, you probably know that it is
[21:27] a total nightbeck. If that resonates, I
[21:30] would love to tell you about Creator
[21:31] Grid, where we will literally run your
[21:33] creator campaigns for you endto end. You
[21:36] don’t pay anything upfront. There is no
[21:38] retainer. There’s no subscription. And
[21:39] you only pay for results.
[21:43] Oh, [laughter]
[21:46] okay. We need a bit of quality control
[21:47] on this one. I I don’t know why it’s
[21:49] bugging out. Like it it looks campaigns
[21:51] for brands like Notion, Whisper Flow,
[21:52] To-Doist, Superhum, Victor, and a bunch
[21:54] of others.
[21:55] >> I see what they’re trying to do, but
[21:56] this is just like super super super
[21:57] buggy. Like, what the hell is it doing?
[21:59] How the hell did you create a video
[22:02] editor within the browser itself? I
[22:05] didn’t build that. The browser is
[22:06] Remotion Studio, which comes free with
[22:08] Remotion. It’s a small local web app.
[22:10] Interesting. Okay, so Remotion Studio is
[22:13] probably the most like similar to a
[22:14] video editing experience, but like it
[22:17] was a bit buggy, just not as good as the
[22:20] one that Hyperframes made. But
[22:21] Hyperframes doesn’t have like the video
[22:23] editing interface. It’s purely the code.
[22:26] And the one in Da Vinci Resolve just was
[22:28] not very good at all. So at this point,
[22:30] I would rate Hyperframes as being like
[22:31] probably a 7 out of 10. Da Vinci
[22:33] Resolve, I would rate it being probably
[22:34] a 3 out of 10. And Remotion was super
[22:37] buggy, so I’d probably give it a 2 out
[22:38] of 10 right now. Uh but it does have the
[22:39] right idea. It does have the right idea.
[22:43] Here is the problem. I think if we think
[22:45] about like the two versions of the
[22:47] video, if we imagine a roll is what what
[22:50] being said that needs to be a cut. So we
[22:53] go for raw a- roll. What we what this
[22:55] kind of needs is a script to raw a-roll
[23:00] then you get the a cut and then you get
[23:02] the add in effects. What I was trying to
[23:05] get Resolve to do is basically take 10
[23:09] GB worth of A-roll footage and hope that
[23:11] it would a magically be able to 45
[23:14] seconds of that to turn into a cut, but
[23:16] then also to actually be able to edit an
[23:19] A cut using J cuts and L cuts to the
[23:21] level that I would like. Now, that is
[23:22] probably a little bit unfair and that’s
[23:23] a little bit beyond what Opus 5.5 can do
[23:26] at this stage. Maybe one day you’ll be
[23:28] able to just give it a bunch of footage
[23:30] and say, “Eedit this for me.” But today
[23:31] is not that day. So, we are going to
[23:33] have to at least on the Resolve front
[23:36] drill down and start doing things a
[23:38] little bit manually. By which I mean
[23:40] kind of breaking things down a little
[23:41] bit more step by step. Maybe that’s the
[23:43] way to think about AI editing with
[23:45] Resolve. Prove the concept first for 30
[23:50] seconds, you know, and okay, this was
[23:52] dumb of me to not realize beforehand. In
[23:53] fairness, it’s always good to just one
[23:54] shot and see what happens because then
[23:55] you see the holes in your [music] in
[23:57] your prompting. Because if we were
[23:58] hiring a human video editor and now
[24:00] we’re testing out a brand new video
[24:01] editor, the way we do that is we give
[24:03] them 10 minutes of footage and we ask
[24:05] them to just give us the first 60
[24:06] seconds as an edit because it gives us a
[24:09] feel for how are they doing. It it would
[24:11] be unfair to ask a human video editor to
[24:13] say, “Hey, of these 10 minutes, pick the
[24:15] best 45 seconds and turn it into a
[24:17] cohesive story and then let’s see see
[24:19] the editing because it would be a lot
[24:20] for a human video editor who’s good at
[24:21] video editing but maybe not good at
[24:23] scripting. Maybe actually the thing that
[24:24] I recorded is not suitable to be spliced
[24:26] into a 45se secondond chunk. Yeah, I
[24:28] don’t want to be one of those people
[24:29] that tries to oneshot something, gets a
[24:31] slightly response, and then thinks AI
[24:33] sucks. Like, I knew it. I knew it was
[24:35] using too much data. I knew the data
[24:37] centers were evil, etc., etc. It’s like,
[24:38] no, I’m sure that there’s definitely a
[24:40] way to do this. I just need to be a
[24:41] little bit more sophisticated with the
[24:43] process and make it happen. So my
[24:46] learning here, learning number one is
[24:48] treat the AI like a human and give it
[24:51] the task that you would give to a human
[24:53] with the same level of instruction.
[24:55] Okay, I’m actually going to switch out
[24:56] of Claude and go into go back into
[24:58] Grock. You know what? Let me just use
[25:00] Swiss flow. So I think what’s happening
[25:01] here is that we actually have two
[25:03] different work streams. One is a pure
[25:04] faceless launch video with like text and
[25:07] music and stuff. And so that is one that
[25:08] doesn’t need the original footage, but
[25:10] then the second one is editing original
[25:12] footage. Those are actually two very
[25:14] different things. Editing original
[25:15] footage involves bringing in the
[25:16] footage, sorting out the A cut, and then
[25:19] doing the B-roll and the animations and
[25:20] stuff. Whereas the pure Faceless is just
[25:23] about doing really good animations. Now
[25:24] I think what we can actually do aha is a
[25:27] hybrid of the two hybrid because for
[25:30] example if we figure out a system of
[25:32] nailing the editing of actual footage
[25:35] like making the a cut really good and
[25:37] then separately we had an absolutely
[25:39] sick way of generating insanely
[25:41] highquality motion video that shows the
[25:45] screenshots of the app with all the
[25:47] relevant zooms and pans and all those
[25:48] things. Those are more similar to the
[25:50] ones that people post on Twitter. We can
[25:52] then combine the best of both worlds to
[25:54] create a talking head explainer video
[25:56] that also has all of the bells and
[25:59] whistles that like a purely animated
[26:01] launch video would have. Yes, treat them
[26:03] as two factories and then a third step
[26:05] that only composites. Stream one, a cut
[26:07] of real footage. Stream two, motion
[26:09] plates, hyperframes when product
[26:10] screenshot, zooms, pans, connected type,
[26:12] music hits, build each plate as its own
[26:14] 3 to 8 second composition against real
[26:15] UI. Drop hyperframes, MP4 into the V3 as
[26:18] B-roll. Talking head stays on V1. Okay.
[26:20] Yep. Nice. Why are you calling them
[26:22] motion plates? Is that like industry
[26:24] terminology for this sort of thing? Oh,
[26:25] it’s just a post-production word for a
[26:27] finished piece of motion graphics that
[26:28] you drop into a timeline. A plate is a
[26:30] self-contained visual layer. Oh, I had
[26:32] no idea. In old film work, it was a
[26:34] painted a photograph background you
[26:35] composite piece people against. In this
[26:36] workflow, it’s a short already rendered
[26:38] movie. 3 to 8 seconds of zoom pans type
[26:39] in product UI that does not need to be
[26:41] built inside Resolve. Ah, motion just
[26:43] means it’s animated. You’re not asking
[26:45] Resolve to animate the dashboard. You’re
[26:46] asking to place a clip that already has
[26:48] the animation in it. That’s clever. One
[26:50] thing I find when learning about AI
[26:51] stuff or when trying to do AI stuff is
[26:53] that I learn terminology that’s actually
[26:56] used in the industry. So, for example,
[26:59] the other day I was doing some
[27:00] thumbnails for a thing. And I learned
[27:01] what a contact sheet is, which is
[27:03] basically like a a series of thumbnails
[27:05] for a storyboard or something like that
[27:06] to explore different concept directions.
[27:08] And then because I then had that
[27:09] terminology, I was able to use that
[27:10] terminology in future prompts. I’ve
[27:12] never heard the phrase motion plate
[27:13] before, but this is exactly what I want.
[27:16] I want the a cut of of the video, the
[27:18] A-roll of me talking, and then I want
[27:20] really high quality motion plates to
[27:22] appear over the top based on what I’m
[27:24] saying in the video. Learning number
[27:25] two, learn the words that the
[27:26] professionals use. Because as soon as I
[27:28] knew it was called a motion plate, I
[27:29] could then ask the AI for exactly what I
[27:31] wanted. All right, this is exciting. I
[27:33] now have more of a mental model of what
[27:36] I’m trying to actually do here.
[27:39] Okay. So, I’m going to start with what I
[27:42] want is I’ll give you a raw a cut a
[27:47] cleaned up a cut of the video of me
[27:51] talking. Ultimately, what I want is
[27:53] really nice motion plates that I can add
[27:57] to the a cut using something like
[28:00] resolve. All right. Hyperframes can do
[28:03] this. Only the motion graphic shows with
[28:05] no blacks blocks around it. PNG
[28:06] sequences with transparency, 4K output,
[28:09] amazing frame rate match to your
[28:11] footage, one plates project with each
[28:12] plate as its own file. Let’s just do
[28:15] exactly
[28:17] the same one
[28:19] for remotion and see. Okay, I think I
[28:22] can basically get rid of worrying about
[28:25] resolve Dinci Resolve for now because I
[28:27] know that I don’t need that for this
[28:29] stage of the process. I’m now realizing
[28:31] that I just need to create the motion
[28:33] plates because then the script, the A-
[28:35] roll, the A cut, that’s like the easy
[28:37] part. I could always do it manually if I
[28:38] really wanted to. I could chuck it into
[28:39] Premiere and use Fire Cut if I really
[28:40] wanted to. The A cut is the easy part.
[28:42] The hard part is what I now know is
[28:44] called the motion plates, which is cool.
[28:45] Ah, it’s rendering a contact sheet.
[28:47] Let’s see. Okay, that’s quite good.
[28:48] That’s quite good. You pick the
[28:49] creators. That is all pretty good. Wow,
[28:51] I like this. The only thing I don’t like
[28:53] about it is the sort of solid colors.
[28:56] Okay, I like this. I just don’t like the
[28:58] solid colors of the backgrounds of the
[29:01] vertical video thumbnails. I want them
[29:04] to look more like actual UGC. So, let’s
[29:07] use real/ AI gen photo realistic images.
[29:12] Oh, wow. That’s pretty good. So, it’s
[29:14] using Canva to generate casual UGC style
[29:18] images to replace this stuff that looks
[29:20] a little bit a little bit fake. I mean,
[29:22] it all looks a bit fake, but like I
[29:23] think that would like sell the the
[29:25] vision a little bit better. Learning
[29:26] number three, get the AI to show you the
[29:28] plan before it builds anything. If you
[29:30] have a storyboard or a contact sheet,
[29:31] that takes seconds to go through and fix
[29:34] stuff. But if you get it to oneshot the
[29:35] render all the way at the end, then like
[29:37] you have to spend a lot more time fixing
[29:38] stuff. All right, so Remotion has
[29:40] actually done a bunch of stuff. Let’s
[29:41] see what it looks like. Inbox triage in
[29:44] 60 seconds.
[29:47] You approve the content. Yep.
[29:54] You pick the creators.
[29:56] Okay, that’s really good.
[30:00] Your campaigns run end to end. That is
[30:03] remarkably good. Well done, Remotion.
[30:07] That’s solid. Look at that.
[30:11] Wow.
[30:14] Okay. I mean, we can definitely make it
[30:16] make a nicer graph, but this is really
[30:19] solid. Okay, I now like reotion again.
[30:21] Let me just give it the same prompts as
[30:24] I did with this one. All right, at this
[30:25] point it’s 3:45 p.m. I have a couple of
[30:27] calls in about an hour and so I want to
[30:29] make as much progress as I possibly can
[30:30] in the next hour. I think we’re getting
[30:32] there. Like the motion plates are
[30:33] actually getting surprisingly good. So I
[30:36] just need a properly scripted 45se
[30:39] secondond video which I can then just
[30:41] record now get I don’t know one of these
[30:44] tools to get the a cut of it. Once I
[30:45] have a clean a cut that I’m happy with,
[30:47] I can then overlay all of these
[30:49] different motion plates that we’re
[30:50] creating whether it’s from reotion or
[30:51] hyperlapse. Okay, so at this point I’m
[30:54] very happy with the images of what this
[30:57] stuff looks like. I think it’s done a
[30:59] really solid job. This is the
[31:00] hyperframes method. Let’s see what it
[31:02] does when it animates it.
[31:06] All right, updated reel. New views
[31:09] chart. Let’s see what this looks like
[31:11] via remotion.
[31:17] Nice.
[31:25] Nice.
[31:28] Okay, well done, Remotion. Now, at this
[31:30] point, the thing I’d finally clocked was
[31:31] that my original video was about 5
[31:33] minutes long. And I’d basically been
[31:34] asking the AI to dig out a 45se
[31:37] secondond launch video out of 5 minutes
[31:38] of me talking, which is a job that I
[31:40] would never give to a human editor
[31:41] either. And so, learning number four is
[31:43] that if you want a short video, film a
[31:44] short script. So, I got Claude to write
[31:46] me a very tight script based on what it
[31:47] said in the original and then I filmed
[31:49] it again. Let’s get Claude to create the
[31:51] script for me because it knows exactly
[31:52] what I said previously and it’s been
[31:54] doing all this stuff. In this project,
[31:55] we should have the transcript of the
[31:58] VSSL I filmed for creator grid. Write me
[32:01] the V1 of the script. Okay, this is
[32:04] actually a very solid script. Let’s now
[32:06] film the thing. If you’re a brand doing
[32:08] creator marketing, you will know that
[32:09] it’s a total nightmare. Okay, so we have
[32:11] just filmed the video. So now we need to
[32:14] export a clean a cut of this video. How
[32:17] do how do I most expediently [music]
[32:20] export a clean a cut? You know what?
[32:23] Let’s download Final Cut for Resolve. I
[32:26] think Final Cut for Resolve, somewhat
[32:28] ironically, is actually the tool that
[32:29] I’m going to use for this. Fire Cut for
[32:31] Da Vinci Resolve. I’ve never actually
[32:32] used Final Cut myself. Our video video
[32:34] editors use it and we built it as an
[32:36] initial tool to use in-house for our for
[32:37] our team. Okay, so now I want to do the
[32:39] first cut. Boom. Can I just generate the
[32:42] first cut? Let’s see what happens. Nice.
[32:45] Okay. If you’re a creator doing paint
[32:47] agency. Yep. I’m a total noob to Resolve
[32:50] having literally never used it before in
[32:51] my life. But let’s see. Let’s see what
[32:53] happens. How do I zoom in and out? Oh,
[32:56] okay. Yeah, fine.
[32:57] >> If you’re a brand doing creator
[32:58] marketing, you will know that it’s a
[32:59] total nightmare. Either you pay an
[33:01] agency a massive retainer and wake
[33:03] either you pay a
[33:04] >> Safari has given me a very rough first
[33:06] cut, but it hasn’t quite made it. You
[33:09] know what? Maybe this is just a case of
[33:11] me actually just doing this manually. Uh
[33:13] what how do I do uh cuts on this thing?
[33:17] Command T. Uh blade tool. Boom. Uh it’s
[33:20] not a magnetic timeline for goodness
[33:22] sake. I haven’t done this sort of in
[33:24] absolutely ages. But this literally used
[33:26] to be my bread and butter in Final Cut
[33:28] >> Produced
[33:30] by us, our team, and our platform.
[33:31] >> So at this point, we have just exported
[33:33] a clean a cut. I probably could have
[33:35] done it with AI, but I didn’t have the
[33:36] time to figure it out, so I don’t know
[33:37] how long it would take me. So now what I
[33:39] want is I want the transcription. So I’m
[33:41] going to use this tool, whisper
[33:42] transcription for that. Okay, we finally
[33:45] have a clean cut that I’ve just edited.
[33:48] This is the transcript. I want you to
[33:51] generate all the motion plates for this
[33:54] new transcript. Boom. And let’s just do
[33:56] the same thing over here. So the A cut
[33:58] is done in Resolve using Far Cut and
[34:00] using a bit of manual editing. Now for
[34:01] the B-roll, motion plates are being done
[34:03] by hyperframes and by remotion. Sick.
[34:06] Then background music. And I think then
[34:09] maybe the final thing might be to add in
[34:12] some sound design actually like plops
[34:16] and whooshes and stuff. Conveniently,
[34:18] Epidemic Sound has a bunch of that stuff
[34:19] in it. They’ve got swooshes, ambience,
[34:21] water, vehicles, whoosh. Can I just give
[34:24] the AI access to Epidemic Sound? Cuz
[34:27] that would be handy if it could find API
[34:31] keys. I love it when there’s API keys.
[34:33] So,
[34:35] I don’t want to do this try and try and
[34:36] do this one shot. We’re going to see
[34:37] what it does with just the motion plates
[34:40] and then we’ll add the background music
[34:41] and the sound effects. But I’ve got it
[34:43] ready so that we can save some time.
[34:44] Bonjour.
[34:46] >> How’s it going?
[34:47] >> Good. You are on the camera.
[34:49] >> Hi.
[34:50] >> Okay, now that we’ve actually done this,
[34:51] while we’re waiting, we can actually
[34:52] check out Parth. What path said, let’s
[34:55] see. If you’re a brand trying to grow
[34:56] with creator,
[34:57] >> that’s really good. [laughter]
[35:00] >> Oh, he gets it,
[35:03] right?
[35:04] >> Marketing, you probably know that it is
[35:05] a total nightmare. Either you pay an
[35:07] agency a massive retainer and then you
[35:09] wait a few months to years for the
[35:11] content that may or may not be usable or
[35:12] you and your team is trying to like
[35:13] manually reach out to creators and then
[35:15] you’re doing the outreach, the
[35:16] negotiation, the invoicing, the review,
[35:17] all of that kind of stuff yourself in
[35:19] house which you just don’t have the time
[35:20] for. If that resonates, I would love to
[35:22] tell you about creator grid where we
[35:23] will literally run your creator
[35:25] campaigns for you end to end. You don’t
[35:27] pay anything up front. There is no
[35:28] retainer. There’s no subscription and
[35:30] you only pay
[35:31] >> Oh, that’s good. It’s [laughter] good.
[35:34] It’s good
[35:35] >> for results. That means if you’re
[35:36] looking for creators for UGC or content
[35:38] for ads, you only pay for the stuff that
[35:40] you approve. Or if you launch an organic
[35:42] CPM campaign, you only pay for the view.
[35:44] >> I mean, yeah, he’s definitely using AI
[35:45] systems as well, but like obviously
[35:47] totally fine.
[35:47] >> Creators actually get you. If we haven’t
[35:48] met, hello, my name is Ollie. I’m a
[35:50] doctor turned entrepreneur and creator
[35:51] with about 8 million followers across
[35:53] socials and I’ve been on the creator
[35:54] side of brand deals for the last decade.
[35:56] But the reason me and my team are
[35:57] building Creator Grid is that for the
[35:58] last 2 or 3 years, we’ve been trying to
[36:00] do creator marketing ourselves for our
[36:01] own software products and realizing how
[36:03] much of a nightmare it is, which is why
[36:04] we built Creator Grid for ourselves. But
[36:05] then as we started speaking to other
[36:07] brands, everyone started saying to us,
[36:08] “Oh my goodness, I need this.” And so we
[36:10] turned it into a product that now lots
[36:12] of big brands around the world are
[36:13] actually using
[36:13] >> nice
[36:14] >> using for their creative marketing. And
[36:15] so in the next few minutes, I want to
[36:16] show you exactly how it works and why I
[36:18] think it is a better way for you to do
[36:20] your creative marketing rather than
[36:21] working with an expensive agency or
[36:23] trying to do it in house. Creator Rit is
[36:24] our attempt to basically give you the
[36:26] level of service that a done for you
[36:28] agency would, just way faster, way
[36:30] cheaper, and with a lot less bandwidth
[36:31] from your team.
[36:32] >> Nice. That’s pretty solid team. The
[36:33] creative platform is kind of like having
[36:35] a teammate who lives in your Slack. So
[36:36] you tell them what you’re trying to
[36:37] achieve and then we will literally run
[36:39] the campaign for you end to end. A bit
[36:40] like how simple it is to run meta ads.
[36:42] You set the objective. So whether you
[36:43] want content for ads or for UGC.
[36:44] >> Bloody hell, that’s good.
[36:45] >> See? Or whether you want an organic CPM
[36:47] campaign and we will handle every aspect
[36:49] of the delivery. All you have to do is
[36:51] basically two things. So firstly, you
[36:52] decide which creators you want to work
[36:53] with.
[36:54] >> Okay. So I mean he’s going going a
[36:56] little bit more like real life rather
[36:57] than like the fake Harry Potter stuff
[36:58] that I was using.
[36:59] >> So we’ll source them for you and you
[37:00] just sort of approve or reject them. And
[37:01] secondly, if you want to approve
[37:03] >> I like the sound design
[37:04] >> content manually. We will give you a
[37:05] bunch of content and you just decide to
[37:07] approve it or not. Everything else is
[37:08] handled by us. So the sourcing of
[37:10] creators, negotiating, reviewing the
[37:11] content initially and then invoicing and
[37:13] paying the creators, tracking all the
[37:14] views, making sure the content is in
[37:16] line with your brand guidelines, all of
[37:17] that annoying stuff we take off of your
[37:19] plate.
[37:19] >> Oh, that’s really good, man. Path is
[37:20] sick of this. Let’s see if we can come
[37:22] close to his level of [music] quality.
[37:24] Yeah, mate. This Mac is absolutely
[37:27] heating up like a like a barbecue.
[37:30] Yeah, actually hard to hard to hard to
[37:33] hold on to it. This means that I can
[37:36] campaign to the wife that I need the
[37:37] fully decked out M5 Max Studio, right?
[37:40] Which costs about 20,000 USD, but it’s
[37:42] worth it. It’s my boy math cuz it will
[37:44] make that AI editing a little bit
[37:45] faster. Even though I don’t really do
[37:47] editing myself. I’m just doing this for
[37:48] the vibes, right? Is that okay? Can I
[37:49] get an M5 Max Studio, please?
[37:51] >> We can talk about it later.
[37:52] >> Nice. Very good. Very good.
[37:55] All right. So, Remotion is rendering all
[37:57] its things. This hyperframes is
[37:58] rendering all its things.
[38:02] Compositing all plates over the cut into
[38:04] a review MP4. That’s what we like to
[38:06] see. Okay, I think we are done. Your
[38:09] first cut with all nine plates dropped
[38:11] in at the time codes for review only.
[38:13] Let’s see what happens. So, this is
[38:15] hyperframes. If you’re a brand doing
[38:17] creator marketing, you will know that
[38:18] it’s a total nightmare. Either you pay
[38:20] loads of money to an agency and wait
[38:21] ages for content, or your own team gets
[38:23] stuck doing everything, finding the
[38:24] creators, DMing them, negotiating with
[38:25] them, doing the invoices, writing
[38:27] briefs, contracts, like all of that
[38:28] stuff. Creator Grit started as our
[38:29] in-house tool for dealing with all this
[38:31] f. But then loads of brands told us they
[38:32] love the idea and so we spun it out. Now
[38:34] we run your creator campaigns end to
[38:36] end, whether it’s UGC for ads or whether
[38:38] you want organic views on content. Okay,
[38:40] that is not bad. I wish it did more
[38:42] picturein picture. Um, okay, not bad.
[38:45] Uh, but I don’t like the full screen
[38:47] nature of the graphics. I want always
[38:51] picturein picture but done in a nice
[38:55] way. Okay, let’s see what Remotion has
[38:58] done. If you’re a brand doing creator
[39:00] marketing, you will know that it’s a
[39:01] total nightmare. Either you pay loads of
[39:02] money to an agency and wait ages for
[39:04] content, or your own team gets stuck
[39:05] doing everything, finding the creators,
[39:07] DMing them, negotiating with them, doing
[39:08] the invoices, writing briefs, contracts,
[39:10] like all of that stuff. Creator Grit
[39:11] started as our in-house tool for dealing
[39:13] with all this f. Loads of brands told us
[39:15] they love the idea and so we spun it
[39:16] out. Now we run your creator campaigns
[39:18] end to end. Whether it’s UGC for ads or
[39:20] whether you want organic views on
[39:21] content. Okay, so at this point we have
[39:24] got V1 of the renders from Hyperframes
[39:27] and from Remotion. Both are like pretty
[39:29] good. I would say they’re like 85%
[39:31] there. There’s a little there’s a bit of
[39:32] polish that they could all do with the
[39:34] one. The Remotion one looks a little bit
[39:35] more AI sloppy with like the gradients
[39:37] and stuff. So maybe we can just like
[39:38] stop it from doing that. And there were
[39:39] a couple of bugs like a few like random
[39:41] frames and like flash frames and things
[39:43] in the reotion one, but the hyperframes
[39:45] one looked pretty good. Okay, so I’m
[39:46] getting both hyperframe and reotion to
[39:48] both do the picturein picture thing that
[39:50] sort of looks like a loom video. All
[39:51] right, while this is still rendering,
[39:53] let’s do a wee. I’ve got 9 minutes until
[39:56] my next call. So maybe while I’m on the
[39:59] call, we’ll get it to keep just cooking
[40:01] in the background for the sound effects
[40:02] and the uh background music. Let’s see
[40:04] if the new Remotion version is good. If
[40:06] you’re a brand doing creator marketing,
[40:08] you will know that it’s a total
[40:09] nightmare. Either you pay loads of money
[40:10] to an agency and wait ages for content,
[40:12] or your own team gets stuck doing
[40:13] everything, finding the creators, DMing
[40:15] them, negotiating, invoices, writing,
[40:17] briefs, contracts, like all of that
[40:18] stuff. Okay, I like the Remotion one. I
[40:20] think we are 90% there now. So, I’m just
[40:22] getting it to tweak the very first line
[40:25] and add a bit more juice to the motion
[40:27] graphics initially. And now I have got a
[40:30] call which I’m going to take on my
[40:32] laptop so I can keep this stuff running
[40:33] in the background without dying. At this
[40:35] point, I was on a call with Pablo, my
[40:37] co-founder on Creator Grade, and he
[40:38] actually sent over a brand new logo.
[40:39] Now, when it comes to normal video
[40:41] editing, doing a rebrand [music]
[40:42] when you’re 90% done with a video is a
[40:44] total pain in the neck because it often
[40:46] means going back and redoing every
[40:47] single graphic by hand. But because
[40:49] these videos were basically code, my
[40:51] theory was I could just give Claude the
[40:53] new logo and I could ask it to update
[40:55] the entire design system based on that
[40:57] logo. And so, let’s see how it does. So,
[40:59] this is the new kind of logo for creator
[41:02] grid. This is the new creator grid logo.
[41:06] Can you update the [music] styles etc of
[41:09] the video to match including the
[41:11] relevant colors and vibe? Boom. And
[41:16] boom. Cool. All right, it is officially
[41:19] 7:05 p.m. I need to head home for family
[41:22] dinner and then baby bedtime with the
[41:25] baby girl. But I think Izzy is still
[41:28] here on some onboarding calls for her
[41:30] new uh Dream 21 community. And so until
[41:33] then, we’re going to see if we can wait
[41:36] for these draft renders to finish and
[41:37] then we’ll see how good a job this does
[41:39] at just like reskinning the videos based
[41:43] on the new logo. Yeah, it’s having to
[41:45] like rerender all of the motion plates.
[41:47] Learning number five, when your video is
[41:49] basically code, changing your mind is
[41:51] super cheap. Imagine how sick it would
[41:52] be if I had a fully decked out Mac
[41:54] Studio that was able to run local AI
[41:55] models and I could just dump footage
[41:58] onto it and overnight it would edit the
[42:01] footage into what is at least a rough
[42:02] cut. And so then our editors just get
[42:04] the rough cut that they can then
[42:05] finalize. That would be absolutely
[42:07] freaking bananas. But for now, I feel
[42:09] like this is a good final test of
[42:12] today’s learning project to learn how to
[42:13] do AI video editing in a day. And here
[42:16] we go.
[42:18] Boss battle.
[42:20] If you’re a brand doing creator
[42:21] marketing, you will know that it’s a
[42:22] total nightmare. Either you pay loads of
[42:24] money to an agency and wait ages for
[42:25] content, or your own team gets stuck
[42:27] doing everything, finding the creators,
[42:28] DMing them, negotiating with them, doing
[42:30] the invoices, writing briefs, contracts,
[42:31] like all of that stuff. Creator Grit
[42:33] started as our in-house tool for dealing
[42:34] with all this f. But then loads of
[42:35] brands told us they love the idea. And
[42:37] so we spun it out. Now we run your
[42:39] creator campaigns end to end. Whether
[42:40] it’s UGC for ads or whether you want
[42:42] organic views on content, you just have
[42:43] to do two things. Firstly, you just
[42:45] approve the creators that you want to
[42:46] work with that we give to you. And
[42:47] secondly, you approve the content that
[42:49] they produce. Everything else is handled
[42:50] by us, our team, and our platform.
[42:52] There’s no retainer, there’s no
[42:53] subscription, and you only pay for the
[42:54] content you approve or the organic views
[42:56] that that content gets. We’re currently
[42:58] working with a bunch of big brands, and
[42:59] we have over 10,000 creators on the
[43:01] platform just waiting to work with you.
[43:02] So, if you’d like to try creator
[43:03] marketing without the chaos, please do
[43:05] get in touch, tell us what you’re after,
[43:06] and we will just send you creators that
[43:07] might be a good fit. It costs you
[43:09] absolutely nothing to find out. Nice.
[43:12] That is really good. It’s background
[43:14] music is a little bit egregious. A few
[43:16] of the sound effects are a little bit
[43:17] like there’s a couple of flash frames
[43:19] where I got things a little bit wrong,
[43:21] but like holy moly, this is like 97%
[43:24] there. Okay, now this was hyperframes.
[43:27] Let’s check out Remotion. If you’re a
[43:30] brand doing creator marketing, you will
[43:31] know that it’s a total nightmare. Either
[43:33] you pay loads of money to an agency and
[43:34] wait ages for content, or your own team
[43:36] gets stuck doing everything, finding the
[43:37] creators, DMing them, negotiating with
[43:39] them, doing the invoices, writing
[43:40] briefs, contracts, like all of that
[43:42] stuff. Creative Grit started as our
[43:43] in-house tool for dealing with all this
[43:44] f. for them. Loads of brands told us
[43:46] they love the idea and so we spun it
[43:47] out. Now we run your creator campaigns
[43:49] end to end. Whether it’s UGC for ads or
[43:51] whether you want organic views on
[43:52] content, you just have to do two things.
[43:54] Firstly, you just approve the creators
[43:55] that you want to work with that we give
[43:56] to you. And secondly, you approve the
[43:58] content that they produce. Everything
[43:59] else is handled by us, our team, and our
[44:01] platform. There’s no retainer. There’s
[44:02] no subscription and you only pay for the
[44:04] content you approve or the organic views
[44:06] that that content gets. We’re currently
[44:07] working with a bunch of big brands and
[44:09] we have over 10,000 creators on the
[44:10] platform just waiting to work with you.
[44:12] If you’d like to try creator marketing
[44:13] without the chaos, please do get in
[44:15] touch. Tell us what you’re after and we
[44:16] will just send you creators that might
[44:17] be a good fit. It costs you absolutely
[44:19] nothing to find out. Wow, that is also
[44:22] not bad. That is also pretty solid. This
[44:24] AI editing thing, holy moly. Do you want
[44:26] to check out what I’ve been cooking all
[44:28] day?
[44:28] >> Yes. Is it filming?
[44:30] >> It’s it’s it’s filming. If you’re a
[44:31] brand doing creator marketing, you know,
[44:33] it’s a total nightmare. Either you pay
[44:34] loads of money to an agency and wait
[44:36] ages for content, or your own team gets
[44:37] stuck doing everything, finding the
[44:39] creators, DMing them, negotiating with
[44:40] them. I mean, these look quite
[44:42] contracts, like all of that stuff.
[44:43] Creator Grid started as our in-house
[44:45] tool for dealing with all this fat, but
[44:46] then loads of brands told us they love
[44:47] the idea. We spun it out. Now, we run
[44:49] your creator campaigns end to end.
[44:51] Whether it’s UGC for ads or whether you
[44:53] want organic music, you just have to do
[44:55] two things. Firstly, you just approve
[44:56] the creators that you want to work with
[44:57] that we give to you. And secondly, you
[44:58] approve the content that they produce.
[45:00] Everything else is handled by us, our
[45:01] team, and our platform.
[45:03] >> Was this one shoted?
[45:04] >> No, of course not.
[45:04] >> Okay. [laughter]
[45:05] >> About 20 shoted.
[45:06] >> I was going to say like this is like
[45:08] this is really good. That’s very good.
[45:10] >> So, if we scroll like this is how much
[45:12] back and forth it was.
[45:13] >> Okay.
[45:14] >> So, it’s it’s it’s not a huge amount
[45:15] like and if we look at the amount of
[45:17] words of you know I’m mostly just giving
[45:19] it minor prompts here and there.
[45:20] >> Yeah. And it’s mostly just sort of
[45:21] >> and it’s mostly just like doing stuff
[45:23] >> doing stuff. Video editors are cooked
[45:25] >> except our guys of course.
[45:26] >> Our guys are great.
[45:27] >> Except our guys of course.
[45:28] >> Yeah. Our guys are using this using this
[45:30] a lot. I saw partit part was also very
[45:32] good.
[45:32] >> Yeah. Um better than this stuff but
[45:34] nice.
[45:35] >> I think human plus AI is the that’s
[45:37] >> the sweet spot.
[45:37] >> It’s the sweet spot right now. Yeah.
[45:40] >> Now, I said at the start of the day that
[45:41] I would be happy with the AI getting to
[45:43] 80% as good as our human editors. By the
[45:46] end of the day, I reckon we got it to
[45:47] about 97%, especially on the motion
[45:49] graphics front. It took about 6 hours
[45:51] from start to finish, and it didn’t cost
[45:52] anything on top of my Claude
[45:54] subscription. And now that I’ve done
[45:55] this once, I’ve firstly got all of the
[45:57] code that Claude generated, and so it’d
[45:58] be easier to do it again. But more
[46:00] importantly, I have more of a mental
[46:01] model for exactly how AI video editing
[46:03] works. I feel like I I have I’ve
[46:05] unlocked a part of my brain that’s now
[46:08] able to think of more ways to use this
[46:11] miraculous technology for loads of other
[46:13] things in the business. Just going to be
[46:15] really fun. And what’s flabbergasting is
[46:18] that this is the worst it is ever going
[46:20] to be. It only gets better from here.
[46:23] Now, if you’ve gotten to this point in
[46:24] the video and you enjoyed it, you might
[46:25] like to check out my video right over
[46:26] here where I talk about tips for using
[46:28] AI to learn anything super fast. That’ll
[46:31] be right over there. Thank you so much
[46:32] for watching and I’ll see