YT YouTube2Text

YouTube2Text → Transcripts → Master AI Video Editing in 25 Minutes (Full Course)

Master AI Video Editing in 25 Minutes (Full Course) — Transcript

by Dan Kieft · 5,558 words · 757 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00A few days ago, I was doing the final
0:01check on one of my videos. Then I
0:03noticed that my editor kept putting in
0:05these kind of disturbing images. So, I
0:08texted him, like, "My hairline isn't
0:10that bad." But, he just sent me more
0:12images mugging my appearance, which made
0:14me wonder, "Do I even need this guy now
0:16that we got GPT Astra?" So, I did some
0:19research and I found six methods of how
0:21you can start using AI to edit [music]
0:23your videos. In fact, everything you've
0:25just seen was all made using AI.
0:29Okay, before we dive into this, I just
0:30want to get one thing clear. The goal is
0:32not to replace my editor. This is just a
0:35joke intro. The idea is, if editing
0:37takes less time, a human can spend more
0:40time on creativity and edit their videos
0:42even better. And that's why in this
0:44video I'll be breaking down all of the
0:46different techniques that go into AI
0:47video editing. I'll show you how you can
0:49easily trim and cut down your video.
0:51Then I'll show you how you can make
0:53multiple different camera angles just
0:55from your A-roll. We will go over
0:56cool-looking motion graphics that are
0:58now easier than ever to create. And I
1:00will even cover a bit of VFX.
1:05Now, last but not least, AI can do a lot
1:07when it comes to sound design. So, make
1:09sure you watch this entire video to the
1:11end, and let's get started.
1:13There are five things that you're going
1:15to need if you want to follow along with
1:17this AI video editing tutorial. The
1:18first one is the obvious one, which is
1:20ChatGPT. We're using Astra as the brain
1:23for all of our projects. The second one
1:25is Higgsfield MCP. This is used for all
1:28of our generations. They're also the
1:30sponsor of this video, but even if they
1:32didn't, this is the best MCP out there,
1:34and they also have an API now, which
1:36we're using for all of the generations
1:39that are about to come that are being
1:40done through GPT Astra together with
1:43Higgsfield. The third thing that you're
1:44going to need, if you're editing through
1:46DaVinci Resolve, you're going to need
1:48the DaVinci MCP. Don't worry, if you're
1:50editing through Premiere Pro, then you
1:52just need Premiere Pro. Same with After
1:54Effects, if you want to do things
1:55through After Effects, you just need to
1:57have that app on your computer. If
1:59you're missing one of the three editing
2:00softwares, as long as you have one of
2:02them, this can still work.
2:05Okay, let's start with basic editing.
2:07We're going to be trimming dead spaces.
2:09For this, you want to download the
2:11ChatGPT desktop app or the Codex app.
2:14The reason why is because the cloud
2:16version, if you're using this cloud, you
2:18have a limit of the file size that you
2:21can attach to this. If you're using this
2:23on your PC, you can click to on your
2:25computer. And now you can upload really
2:27big files that are already on your
2:29computer and it can use it directly from
2:31that and work directly from that. In the
2:33cloud, as you can see, you're limited up
2:35to but my camera files are way bigger
2:37than that. So, make sure you switch to
2:39on your computer, then you drop in the
2:41video that you want to edit. And here
2:42you can select GPT Astra. I'm using
2:45medium most of the times
2:49because it's already expensive enough.
2:50You could go high, but only if you're
2:52rich. Then, we need to start prompting
2:54it. And I have a very detailed prompt
2:56that anyone can use. It's a universal
2:58prompt. I'll explain to you right now.
3:00So, you upload your video and then
3:02you're going to add in this exact
3:04prompt. And this is a super long prompt.
3:06So, I'm going to break it down for you.
3:08Basically, what this prompt does is it
3:09first instructs ChatGPT to trim the raw
3:12footage. Then, it will ask it to
3:14transcribe that footage and trim out the
3:16dead spaces. After that, it will add in
3:18editor instructions as adjustment
3:20layers. Then, it will figure out
3:22potential spaces in your video where you
3:24can have or add in B-cam or even C-cam
3:27and it uses that with the markers. And
3:29lastly, it will give you the finished
3:31version in Premiere
3:35Pro or DaVinci Resolve. I'm using
3:37Premiere Pro right here because we are
3:39using that for most of the editing that
3:41we do on this channel. And this is what
3:43my team is also using. But this also
3:45works with DaVinci. I will show you an
3:47example later. You might get some things
3:49that you need to allow. So, I'm just
3:51going to do always allow for now. This
3:52might cost you more slow. You can
3:54literally see it take over your PC and
3:57clicking on things for you. Still, some
3:59of these things annoy me as it took like
4:0114 minutes to do some of these trims.
4:03So, I would only do this while you're
4:05working on something else and while you
4:06don't want to do the trimming yourself
4:08because you still got to sometimes
4:09monitor if it's doing everything
4:11correctly. So, it's literally taking
4:13over your PC and it's doing everything
4:15for you. As you can see, we now have a
4:17new version where it's marking potential
4:20B-roll. It is adding some adjustment
4:23layers. It is doing everything for me.
4:25Like it's now even rendering. Now, as
4:27you can see, ChatGPT has run all of
4:29these commands using my Premiere Pro and
4:32it made this version. I actually stopped
4:34this because I already had this version
4:36to begin with and that's the one I want
4:37to show you. Because here it just
4:39basically has done the same thing. We're
4:41just one step ahead, which I'm going to
4:42explain to you in a second. But, I
4:44always start my videos with a clap to
4:46sync the audio. Okay, this is the intro.
4:48A few days ago, I was doing the and
4:50using these adjustment layers, it says
4:52intro start here. Then other notes that
4:55I have, like for example here I say
4:57something about I want to add in some V
4:58effects here and then like it mentions
5:01it. Trying to do this, I'm not sure if
5:02this works. I'm trying to think like
5:04what if we can Like this is not the part
5:05that I want to have in the video. These
5:07are just instructions for me and my team
5:09to go over later. So, we can manually
5:11trim these out or you can even ask
5:13ChatGPT to do that for you. The other
5:14thing that it has done is that has
5:16marked these potential B cam positions
5:18and the potential or optional C cam
5:21positions and here again optional C cam.
5:25Let me explain to you what I mean by B
5:27cam and C cam. So, this what you're
5:29watching right now is the A cam.
5:31Then this is a B cam. It's a secondary
5:34side angle. I don't have a camera there.
5:37Like there's no camera there. I wish
5:39there was a camera there. If you have
5:40like three different cameras, like
5:41that's expensive. It's a lot of storage.
5:43Storage these days is expensive. But, we
5:45can use AI. So, this clip is entirely
5:48generated with AI, and I literally asked
5:51Astra to do that for me. Now, the other
5:52thing that we can do is we can also
5:55create C cam. That's another angle. You
5:57can literally create as many different
5:59angles as you want. So, how do you add
6:01in multiple camera angles? Well, the
6:02first thing you need is you need to have
6:04a reference that you can feed ChatGPT to
6:06what your surrounding looks like. So, I
6:08gave it this image. It's with all of the
6:10studio lights turned off. It's a bit of
6:12a behind-the-scenes. It's It's not the
6:13most clean-looking setup. Like, what the
6:16hell's happening here? I'm not going to
6:18I'm I'm renting this place, okay? I
6:20cannot do too much to the place. But, I
6:21then use the same ChatGPT text. But,
6:24this time we're going to use this entire
6:26prompt. What it does is we're going to
6:28ask it to add in the B-roll on specific
6:32segments. So, I'm just going to copy
6:33this prompt. We're going to go back to
6:35ChatGPT, and now we're going to enter it
6:37in there. And now, it you can just ask
6:39it, and it will start up Premiere Pro
6:42again, and it will start generating all
6:44of these things if you have the
6:46Higxfield MCP connected. That's the main
6:48thing you need to do. So, you're
6:49basically going to start up this new
6:50chat by pasting in that prompt. Then,
6:53you're going to add in that extra
6:54reference of your room, of the space, of
6:57the side angle. And then, you're going
6:59to say, "Use the Higxfield MCP." Make
7:02sure you have the Higxfield MCP
7:04connected. Basically, what you can ask
7:06ChatGPT to do, I will show you right
7:08now. Install Higxfield MCP. Install
7:11DaVinci MCP. While you're at it, you
7:13might as well install Blender MCP. And
7:16literally, by just sending this, it will
7:18install all of these things for you if
7:20you have a subscription. If you don't
7:21have a subscription, get a subscription
7:23first. If you want to support my
7:24channel, use my link. Basically, this is
7:26the task that you need to do. Okay, once
7:28you've done that, you can hit send. And
7:30now, it will get to work using that
7:32prompt. Again, this process might take
7:35some time as it's still quite slow, and
7:37we don't have an actual Premiere Pro
7:39plugin [music] yet. With DaVinci, this
7:41might be a lot quicker. But, now it's
7:42also like generating these videos using
7:45Seedance 2.5. And again, like that's a
7:47lengthy process regardless. Now, you can
7:49see it working here right now. So, you
7:52can see it use these Higfield MCPs. And
7:54if you then go over to Higfield itself
7:56and you go to the video tab, you can
7:58literally see it prompt this. And it is
8:00using Seedance 2.5. It's asking it to
8:03replace video one using this as a
8:04reference. Pretty sick, right? I'm just
8:06going to tell you like it might stop
8:08occasionally because right here it's for
8:10example, waiting on the video to
8:12confirm. But, this is the B-roll that it
8:14made for that video. Then I noticed that
8:17my editor kept putting in these kind of
8:20disturbing images. Keep in mind, this is
8:22entirely like AI-generated. I didn't
8:24record this myself. It's just using it
8:26based on this reference. And it made up
8:28itself what I look like from the side.
8:30And looking at everything, it's done
8:32quite an amazing job. Like we can even
8:34see like you don't see it, but I see
8:36there a shadow on my wall, like this
8:38shadow. Then if we look at the C-cam,
8:40even this one. It's time a human can
8:43spend more time on creativity. Like it's
8:45really, really good. Like this thing
8:47there,
8:47>> [music]
8:48>> that light,
8:49actually is positioned like that. The
8:50light behind me is positioned like that.
8:52Just this mouse mat thing is not
8:54positioned like I don't have that. But,
8:55overall, it's quite good. The only thing
8:57is this is if you want to have C-cam. I
8:59don't think this this angle is
9:00flattering on me. So, I just want to use
9:02the B-cam and the A-cam. You can also do
9:05this in DaVinci because it is much, much
9:07quicker as someone has created a MCP for
9:10DaVinci. So, basically, by just
9:12prompting it, use the DaVinci MCP,
9:15within a heartbeat, you would have a
9:17trimmed version of your timeline. So, if
9:19you are a DaVinci editor, or if you are
9:21trying to get into editing yourself, you
9:23might actually want to try out DaVinci.
9:25Keep in mind, you need to use the paid
9:27version for this as the free version
9:29doesn't have the MCP to my knowledge.
9:31But, this is to show you how you can
9:33generate multiple different angles of
9:35your main A-roll. All right. So, you now
9:37understand how you can trim up your
9:38video and how you can create multiple
9:41different angles of the person that's
9:43talking on the screen. But, the next
9:44thing that I'm going to show you is
9:46B-rolls. B-rolls are a great way to
9:48spice up your videos. The problem,
9:50however, is that generic stock footage
9:52B-roll looks boring and might not fit
9:55your vibe. The solution is generating
9:58those B-rolls yourself using AI. You can
10:00completely customize them based on the
10:02prompt you put in. Anything is possible,
10:04really. The other thing is is that you
10:06can get exactly what you want in like 10
10:08minutes. You can save credits by using
10:10cost-efficient models like Kling and
10:12they still do the job quite well.
10:15Cool. So, how do you generate the
10:16B-rolls like I've just shown you? The
10:18first thing you're going to need is you
10:20need to have multiple different
10:21references of the character that you
10:23want to have B-rolls from. In this case,
10:25it's me. So, what I did is I used my
10:27iPhone and then I made these three
10:29images to get a full picture of my face.
10:32But, if you have something a bit more
10:33complex, then you might want to use a
10:36character sheet like this one or this
10:39one. This one is like 100% accurate like
10:42I literally scanned my body and this is
10:44actually how muscled I am. I'm just
10:46trying to make a joke here cuz I'm not
10:48that muscled. But, as you can see, this
10:50is super simple. I used my face as a
10:51reference, then I just added in a prompt
10:54explaining what I want to have on the
10:56screen. If you want to see all of the
10:58prompts that I'm using in this video,
11:00then make sure you're joining the free
11:01School community. Here, for every video
11:03that I make, I'm uploading all of the
11:05assets, all of the prompts, sharing them
11:07with you so you can go through that at
11:09your own pace. To create your B-roll, I
11:11will just give you a quick run down what
11:13I did is I uploaded the three images of
11:16me, then I used this skill MD file that
11:19explains how to make a character sheet,
11:21and then I just gave it a simple prompt
11:22like this. Using the skill and my three
11:24reference images, can you help me
11:26generate a character sheet in
11:28Higgsfield? You also want to might add
11:30in Higgsfield like this and then explain
11:34what you want it to look like. So now
11:36you hit generate. After it's done
11:38generating, it will give you the
11:39character sheet that would look
11:40something like this. You now are ready
11:42to generate your video. But first, you
11:44need to have a video that needs B-roll.
11:46So I recorded this segment for that.
11:50>> [snorts]
11:50>> So the sentence is, if you're stuck for
11:51ideas on what to generate next, go for a
11:53walk or take a shower. Sometimes the
11:55best ideas come when you stop trying so
11:57hard. If you're stuck for ideas and you
11:59don't know what to generate next, the
12:01best thing that you could do is you can
12:02either go for a walk or take a shower
12:05because sometimes the best ideas come
12:07when you stop trying so hard.
12:10Okay.
12:10So at the part of go for a walk, I want
12:12to have B-roll and at the part of taking
12:14a shower, I want to have B-roll just for
12:17visual context. So at the part where I
12:19say you go for a walk, I want to
12:20generate a B-roll for that. So I'm
12:22basically asking it to generate me with
12:25this character sheet, a Cidence 2.5
12:28video of me walking in the park. Then I
12:30describe what my shot looks like. I do
12:32the same for the shower part. Then
12:34literally what it does is it uses
12:36Hicksfield again to generate the video
12:38and it will give you the video right
12:40here. And then if you combine that, you
12:42get something like this. If you're stuck
12:44for ideas and you don't know what to
12:45generate next, the best thing that you
12:47could do is you can either go for a walk
12:49or take a shower because sometimes the
12:51best ideas come when you stop trying so
12:54hard. Okay, let's go a little bit more
12:56advanced. Let's talk about motion
12:58graphics. Here's an example of the
13:00motion graphics that AI made for me.
13:02Video editors need to start using GPT
13:05Astra ASAP. Not only does it create
13:07small motion graphic elements like this,
13:10it can also create a full range
13:12animations that used to take hours in
13:14After Effects. All of that on autopilot.
13:17If you want to learn how to use GPT
13:18Astra, then check out the full tutorial
13:21that is in my free school community
13:23linked down below. Here's the entire
13:25timeline that AI helped me make for all
13:29of these. And I'm naming this a bit more
13:31advanced because the misconception is AI
13:34is this magical thing that will do
13:36everything for you and it will look
13:37amazing. You won't get these results by
13:40just prompting it. You need to combine a
13:42bit of your own skill together with AI
13:44and that's why I'm saying like editors
13:46are not replaced. Like me and my editor
13:48actually collab on this. So we're trying
13:50to figure out what can be done and what
13:52cannot be done with AI. And we have used
13:55of AI. It was also a bit of a painful
13:56process sometimes to figure out how to
13:58do these different style of motion
14:01graphics. I'm going to show you three
14:03different workflows. The first one is
14:04this basic animation workflow. What I
14:07did is I uploaded references of what my
14:09community looks like because I would
14:10love to have some cool looking motion
14:12graphics of that community. Here's the
14:14banner of that community. Still need to
14:16update this. Then I also added in some
14:18style inspiration. So I found these
14:21different type of motion graphic
14:23inspiration. And this is kind of like
14:25what I screenshot it and then I asked it
14:27to help me. Then I basically used
14:29whisper flow to give me these motion
14:31graphics and it just went to work. Like
14:34it just used After Effects and it made
14:36all of these things for me. And after
14:37going a bit of back and forth, it
14:39finished the After Effects motion
14:41graphic for me. It even gave me the
14:43complete project file so I can go in
14:45there and I can edit it myself and I
14:47have the finished video. So again, I am
14:50not an After Effects user myself, but my
14:52editor is. So I can now give this to my
14:54editor to tweak little things. The next
14:56one is more of an advanced workflow and
14:58this is quite hard to explain because
15:00there are so many different prompts that
15:01went into this, but I will try my best
15:03to give you a summary. First, we're
15:04talking about this kind of motion
15:06graphics and this is cool because we're
15:08trying to create a new style that would
15:10add to my videos where we have custom
15:12animated images of me that we can use in
15:14different situations. The way I made
15:16this is I added in a bunch of different
15:18references of styles that I liked and
15:20then again, I just started talking using
15:22Whisper flow about the kind of
15:24animations and things that I want to
15:26see. So then it started to create these
15:28references of me first, and then
15:30eventually it made these images of GPT-
15:333 6 and me again, and then we got this
15:36generation. So it made like all of these
15:38different images and it kept me updated,
15:40and then it gave me the prompts of what
15:42it wanted to use. And then I manually
15:44used these prompts inside of Higgsfield
15:46because I didn't trust Astra and I
15:48didn't want to waste too many credits.
15:50Keep in mind, if you're doing this and
15:51you have the approve for me or full
15:54access available, it might cost you a
15:56lot of credits if you go into the wrong
15:57direction. If you have money, you're
15:59good. If you don't have money, please
16:00don't do it. Like I'm not doing this
16:02either. Like this is risky. Like I
16:03always want to approve because you will
16:05burn through tokens so [music]
16:07quickly using this method. So then I got
16:09like all of these different prompts. I
16:11manually created them inside of
16:13Higgsfield, and then I tweaked like all
16:14of the different settings. So this is me
16:16going back and forth before I did the
16:18final generation. And then eventually
16:20like this is the whole chat. Like you
16:21can see how long it is. Eventually like
16:23it made like all of these different
16:25things for me, and then I imported them
16:27into my timeline and I added like a
16:29small little things like this white
16:31transition right here. But all of this
16:33is done hand in hand with Astra. Okay,
16:36so there's one motion graphic, which is
16:38this pixel version of me, and I'm going
16:40to share with you how I made this
16:42because I think this is awesome for
16:43adding in this into your video. So I
16:45just uploaded an image of me and I just
16:47told him like this is a screenshot of me
16:49and I asked it to create a pixel version
16:51of me. So it made like these three
16:52different versions using Higgsfield MCP.
16:55Then I basically asked it if it can now
16:57give me a Cidens 2.5 prompt and I wanted
17:00to do this part manually again. [music]
17:02So now it gave me the prompt where I
17:04basically asked it what I wanted to have
17:06happen. So in this case, I wanted it to
17:08walk into the frame and say don't forget
17:11to subscribe. So now if you upload the
17:13part where you want to have that in your
17:14video together with your asset and then
17:17you ask it to replace it, and we get
17:19this cool-looking result where I walk
17:21into the frame and say, "Don't forget to
17:23to subscribe." and I jump out of it. Not
17:25only does it create small motion graphic
17:27elements like this,
17:29so now I can use this asset over and
17:31over and over again in all of my videos.
17:34One of my favorite things to do with AI
17:35is making VFX. So now I'm going to show
17:37you how you can do an effect like this.
17:40So you want to learn how to do VFX.
17:42Well, let me actually
17:45Let me show you. Wait.
17:47What the hell is that thing behind me?
17:52Okay, and for my next trick, let me
17:54actually move back a little bit.
17:56So, now we're going to do this.
18:02Oopsy.
18:05So there are a couple steps that go into
18:07making a video like that. First of all,
18:09it starts with your recording. Now this
18:11was my raw recording, and I already
18:13envisioned what I wanted to do. So I
18:15wanted to have something like there.
18:17Then I wanted to use some kind of like
18:19magic wand. In my case, it's actually a
18:21pen that I hold in my hand. And then
18:23something needed to happen at the end.
18:26Let me actually move back a little
18:27[music] bit.
18:28So, now we're going to do this.
18:34Okay, I hope that works.
18:36Yeah, so I I hope that works. So I've
18:39identified three points. First point
18:41being is when I look over my shoulder.
18:43So I'm going to leave a mark right here.
18:45An important thing I want you to do
18:46right now is if you're using Premiere or
18:49if you're using DaVinci and your
18:50timeline looks like this where you can
18:52just see like normal time of how many
18:55seconds it's into the video, change this
18:57to the view. I do that with command and
18:59then clicking on this. Probably would be
19:01control if you're on Windows. And then
19:03you can see the exact frames. So at the
19:05107 frame, I'm going to leave a mark,
19:07and remember that frame for when you're
19:09prompting. Then the next thing is here
19:12when the pen arrives in my hands, so I'm
19:14just going to leave a mark right here at
19:16279. So, I'm just going to add in a
19:18mark, and then I'm going to do the same
19:20when I'm clapping into my hands. So,
19:22right here I want something to happen
19:24with like a car going through the wall.
19:26That is the first things that you need
19:27to do. So, now you have identified these
19:29three points of where the effects need
19:31to happen. Then we're going to walk
19:32through them step by step. So, the first
19:34thing we need is a creature. My first
19:37version of the creature was this. I made
19:39a mistake here where this didn't really
19:42look photorealistic. Until after a few
19:45generations, I finally landed on
19:47something like this. This looks a lot
19:48more creepy, bit more realistic, and
19:50then this ended up being the creature
19:52that I wanted to use. The second thing
19:54that we need to generate is the magic
19:56wand. So, here I made a magic [music]
19:58wand like this. Again, the prompt is
20:01simple. I'm just describing like what I
20:03want to make, what the background looks
20:05like, the lighting, the style, and I'm
20:07using Seedring 5.0 Pro for this. So, now
20:10we already have two over things. Then
20:12the last thing is the car. And here I
20:14have this guy. I call him Bruiser, and I
20:17wanted him to be in a Tesla. Bruiser is
20:20going to run into me, and this is the
20:22reference that we're going to use for
20:23it. So, we now have created the three of
20:26our references. Then what you can do is
20:28you can take all three of these
20:30references, put them inside of ChatGPT
20:32or Claude, then you add in your video,
20:35and then you're going to precisely
20:36describe what you want to have happen
20:39per frame. So, as we remembered, I have
20:42here frame 107, then I have frame 279,
20:47and then I have frame 567. So, that is
20:49what we're going to prompt, and if we do
20:51that correctly, then it will give you a
20:53very detailed prompt of everything
20:56that's happening precisely on the
20:58timestamp. Like it will come up with
21:00that timestamp, and it will tell you
21:01exactly where that thing needs to go,
21:04and it will reference these images. Like
21:06this deserves like a a video on its own,
21:08but I just want to show you like what's
21:10possible. After doing that, you can look
21:12at it side by side, and then you have a
21:14result like this. So, you want to learn
21:17how to do VFX? Well, let me actually
21:21Let me show you. Wait.
21:23What the hell is that thing behind me?
21:28Okay, and for my next trick, let me
21:30actually move back a little bit.
21:32So, now we're going to do this.
21:38>> Oopsie.
21:40>> So, that is exactly how you can generate
21:42these VFX. If I went too fast, please
21:44let me know. I can make a separate video
21:47about this if you want to.
21:50Moving on, I will show you how you can
21:52use Astra 6 to color grade your footage.
21:55You can do this both through ChatGPT
21:57itself, and also through DaVinci. Let's
21:59first start by doing this in Astra 6.
22:02Basically, what you can do is I have
22:03this footage. This was shot while I was
22:06in Malaysia. This is S-Log footage. I'm
22:10using the official Sony conversion LUT
22:12that I found online for my camera. And
22:14then basically, I'm giving it a simple
22:16prompt where I'm saying that the
22:17uploaded MP4 file is my S-Log 3 footage.
22:21And the other file is the official Sony
22:23recording 709 conversion LUT. I want to
22:26use this to color grade my MP4 by using
22:28the conversion LUT. And please, for the
22:30deliverables, can you put this in a
22:32Premiere Pro timeline so I can manually
22:35edit it further if I need. That's all I
22:38do. So, now you hit send, and then it
22:40will do everything for you. So, it has
22:42now given me this folder, and if we look
22:45at these side by side, then you can see
22:47that we now have a color graded video.
22:49This is the super simple way. Again,
22:51sometimes this might take a while. In
22:53this case, you don't want it. I
22:55specifically told it, "Don't use my
22:57Premiere Pro directly, because that's
22:58the slow version." Like for more tasks
23:00like these, you might not have to use
23:03AI. If you have more advanced ones, then
23:05I would definitely use AI, but keep in
23:07mind like it can still be quite slow.
23:09The other thing you can do is you can
23:10use the DaVinci MCP. If you have the
23:13premium version of DaVinci, then you can
23:16ask it to color grade directly in
23:18DaVinci. It is much quicker and it will
23:20also give you much more customization
23:23options because DaVinci is best for
23:25color grading. You can even ask it to be
23:27creative when it comes to color grading
23:29your work. The last segment, which is
23:31also a very important segment, is sound
23:34design. Sound design is what sells your
23:37videos. If you're doing sound design
23:38manually, you can now use AI to do sound
23:42design for you. There are various
23:43different ways. And again, this also
23:45might deserve a video on its own because
23:46you can literally have Astra 6 control
23:49your sound design software, which is
23:51crazy on itself. But I'm going to show
23:52you a simple version. Remember when we
23:54had this video?
24:01>> [music]
24:04>> I personally don't really like the
24:06sounds that [music] Cedance made for
24:08this one. So, what we can do is we can
24:10upload this into Astra, and then we can
24:12just give it a super simple prompt. I
24:14want you to craft a Cedance 2.5 prompt
24:16to generate an audible synchronized
24:18sound effects track for this 8-second UI
24:21animation. The reference is the video,
24:22so you need to analyze it frame by frame
24:24and understand what sound effects. Keep
24:26in mind, you can do this for any type of
24:27video. In this case, I used this
24:29animation, but you can also literally do
24:30this for a intro video. That's how we do
24:33it sometimes. Then you just going to ask
24:35it to make a prompt. In this case, I'm
24:37going to make the prompt because I want
24:39to save some credits before it does this
24:41in Hixfield and I get a version that I'm
24:44not happy with because I always want to
24:46check the prompts that it's going to
24:47use. So, as you can see, it wrote a
24:49super detailed prompt of the audio. And
24:52now you want to go through this and see
24:54if you agree with this. If you disagree,
24:56then go back, reiterate, change a few
24:58things. Always do that with your
24:59prompts. Just don't blindly trust
25:01ChatGPT for everything, because that's
25:03how you're going to burn a lot of money.
25:04Like, trust me. I've been there. If you
25:06have any changes, change them.
25:08Otherwise, copy this, go into Hix Field,
25:11upload your original track in Cines 2.5,
25:13then paste in the prompt right here.
25:15Make sure you add in video one to refer
25:17to that video. Then match the length of
25:20the video, choose your settings. You
25:22could do this in low settings. You don't
25:24need to do this in high settings,
25:25because I'm going to show you why. Here
25:28I have the finished result of that
25:30prompt.
25:37Like, it sounds a bit better, but the
25:39video quality has noticeably gotten
25:41worse. So, you just want to download
25:42that video, then you want to separate
25:44those audio tracks, and then you want to
25:45replace that with the original video.
25:47Literally, you could even drop in that
25:49video in there, and you can just use
25:52like Whisper Flow, and you can say
25:53something like, "I want you to replace
25:56this audio of the file that I just
25:58uploaded, and replace it with the
26:00original file that I gave you, and make
26:02it match." Then, Astra will do the work
26:05for you, and you will have [music]
26:07your original high-quality clip with the
26:10new audio that you just generated. All
26:12right, that is the complete guide of how
26:15you can edit your videos using AI. In
26:17particularly, we're using Astra 6 as the
26:20brains of this. If you want to try this
26:22out yourself, I would recommend using
26:24Hix Field for that. I will leave the
26:25link in the description down below. And,
26:27if you have any questions on this video,
26:30you want to see more things in depth.
26:32Lastly, make sure to click the video on
26:33the screen right now if you want to
26:35learn what else you can do with Astra 6
26:37and Cines.

About this transcript

This page contains the full transcript of Master AI Video Editing in 25 Minutes (Full Course) by Dan Kieft, generated from the public captions YouTube serves with the video. The transcript has 5,558 words across 757 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.