WEBVTT 00:00:00.000 --> 00:00:03.879 So if you're trying to scale video content today, 00:00:04.320 --> 00:00:07.839 you've probably run head first into this conflict. 00:00:08.619 --> 00:00:12.640 You need these professional, realistic AI avatars, 00:00:12.939 --> 00:00:15.199 maybe for marketing, maybe explainer videos, 00:00:15.640 --> 00:00:18.620 but... Every platform seems like every single 00:00:18.620 --> 00:00:20.699 one traps you pretty quickly. Yeah, it's the 00:00:20.699 --> 00:00:23.100 extensive subscriptions, right? Yeah, and the 00:00:23.100 --> 00:00:24.940 the credit limits they get you every time you 00:00:24.940 --> 00:00:27.039 feel kind of locked in exactly and you realize 00:00:27.039 --> 00:00:29.600 Wow, if I want to make say 50 videos this month, 00:00:29.620 --> 00:00:31.839 you're gonna need a second mortgage Yeah, it's 00:00:31.839 --> 00:00:34.219 nuts. But you know the core idea from the sources 00:00:34.219 --> 00:00:36.140 you send over it's actually brilliant we can 00:00:36.140 --> 00:00:39.079 like Completely sidestep that whole system right 00:00:39.079 --> 00:00:41.359 and the promise here. It isn't just about getting 00:00:41.359 --> 00:00:44.140 better avatars. It's about getting Well, unlimited 00:00:44.140 --> 00:00:46.159 creative freedom and really cutting down the 00:00:46.159 --> 00:00:48.759 costs. That's it. So today we're diving into 00:00:48.759 --> 00:00:51.979 this custom three tool blueprint. The idea is 00:00:51.979 --> 00:00:54.500 it turns this expensive limited process into 00:00:54.500 --> 00:00:57.560 something faster, like an asset creation machine. 00:00:57.770 --> 00:00:59.509 Yeah, that's the mission today. We're going to 00:00:59.509 --> 00:01:02.270 unpack this workflow. It really focuses on consistency 00:01:02.270 --> 00:01:05.829 first using a tool called NanoBanana, then smart 00:01:05.829 --> 00:01:08.870 scripting using ChatGPT, and then finally bringing 00:01:08.870 --> 00:01:11.489 in the heavy hitter for realism, that's Hagen, 00:01:11.909 --> 00:01:14.950 but specifically with Google's VEO 3 .1 model 00:01:14.950 --> 00:01:17.489 integrated. All right, let's unpack this then, 00:01:17.569 --> 00:01:20.230 starting with the problem, the bottleneck, financially 00:01:20.230 --> 00:01:24.040 and creatively. Why does just using one platform, 00:01:24.079 --> 00:01:26.340 let's say, sticking only with Hagen, why does 00:01:26.340 --> 00:01:29.299 that fail if your goal is serious scaled content? 00:01:30.100 --> 00:01:32.920 Well, the single tool problem, it's like you're 00:01:32.920 --> 00:01:34.819 paying the vendor for two really different things 00:01:34.819 --> 00:01:36.799 at the same time. Okay. You're paying them for 00:01:36.799 --> 00:01:39.099 the creation part, which is actually pretty costly 00:01:39.099 --> 00:01:41.099 getting the avatar just right, the outfit, the 00:01:41.099 --> 00:01:43.060 look. And then you're also paying them for the 00:01:43.060 --> 00:01:46.060 final video render, that creation part. It just 00:01:46.060 --> 00:01:48.099 burns through credits like crazy. And I guess 00:01:48.099 --> 00:01:50.180 if you need just a small change, like the avatar 00:01:50.180 --> 00:01:51.900 needs a blue shirt today instead of yesterday's 00:01:51.900 --> 00:01:54.129 gray one. Exactly. You're basically paying for 00:01:54.129 --> 00:01:56.030 creation all over again. You're stuck with their 00:01:56.030 --> 00:01:58.469 options, their internal library. Any change means 00:01:58.469 --> 00:02:00.609 another expensive credit burn. The whole trick 00:02:00.609 --> 00:02:03.989 is to sort of decouple those two steps. Ah, so 00:02:03.989 --> 00:02:06.430 the solution is don't start building the avatar 00:02:06.430 --> 00:02:08.310 from scratch inside the expensive video app. 00:02:08.909 --> 00:02:11.310 You pre -make it, pre -customize it somewhere 00:02:11.310 --> 00:02:13.969 else first. Yeah, use cheaper tools or even free 00:02:13.969 --> 00:02:16.449 ones to get the look right. Then you just import 00:02:16.449 --> 00:02:18.909 the finished asset. It's a shift in where the 00:02:18.909 --> 00:02:21.550 cost happens. It's just smart economics, really. 00:02:21.740 --> 00:02:23.840 And that brings us right to the three tools. 00:02:24.400 --> 00:02:27.219 NanoBanana is the base, the foundation. It makes 00:02:27.219 --> 00:02:29.159 that consistent face, but lets you change everything 00:02:29.159 --> 00:02:32.199 else. Clothes, backgrounds, emotions. It locks 00:02:32.199 --> 00:02:34.860 the identity. Okay, tool one, identity lock. 00:02:35.000 --> 00:02:38.560 Now tool two is ChatGPT. You call it the strategist. 00:02:39.240 --> 00:02:41.659 Or like the creative director. It helps you generate 00:02:41.659 --> 00:02:44.300 the perfect detailed instructions, the props 00:02:44.300 --> 00:02:46.620 for NanoBanana. And crucially, it helps you write 00:02:46.620 --> 00:02:48.520 conversational scripts. Scripts that actually 00:02:48.520 --> 00:02:51.240 sound human. which is harder than it sounds. 00:02:51.620 --> 00:02:53.520 And the final piece, the heavy lifting for the 00:02:53.520 --> 00:02:55.659 video itself, that's left to Hagen. Yep, hey 00:02:55.659 --> 00:02:58.870 Jen, but only for the final video render. You 00:02:58.870 --> 00:03:01.150 feed it that pre -made image from NanoBanana, 00:03:01.490 --> 00:03:05.189 and VEO 3 .1 turns that static picture into this 00:03:05.189 --> 00:03:08.469 really lifelike video. But you skip the big creation 00:03:08.469 --> 00:03:11.129 fee inside Hagen itself. OK, this makes sense. 00:03:11.189 --> 00:03:13.050 The whole system seems to rely on separating 00:03:13.050 --> 00:03:15.990 that cost. So why is the order of using these 00:03:15.990 --> 00:03:19.210 three tools so critical? Why does it save money 00:03:19.210 --> 00:03:21.669 and give more freedom? The order saves money 00:03:21.669 --> 00:03:24.050 because creation happens before expensive video 00:03:24.050 --> 00:03:25.889 rendering costs are applied, right? You front 00:03:25.889 --> 00:03:28.370 load the free creation work. Got it. So step 00:03:28.370 --> 00:03:31.409 one then nano banana you called it magic What 00:03:31.409 --> 00:03:33.930 exactly is this thing and how does it fix that 00:03:33.930 --> 00:03:36.250 big avatar problem the face changing all the 00:03:36.250 --> 00:03:39.129 time? So nano banana, it's a pretty powerful 00:03:39.129 --> 00:03:40.949 Google image model right now You can access it 00:03:40.949 --> 00:03:43.250 for free inside Google AI studio, which is like 00:03:43.250 --> 00:03:45.930 their testing area Okay, and it's killer feature. 00:03:45.949 --> 00:03:47.770 The thing that makes it special is character 00:03:47.770 --> 00:03:50.449 consistency character consistency That sounds 00:03:50.449 --> 00:03:52.949 a bit like jargon maybe but it's the fix for 00:03:52.949 --> 00:03:54.810 that image drift right where the face subtly 00:03:54.810 --> 00:03:58.270 changes exactly It means keeping the person's 00:03:58.270 --> 00:04:00.909 face, their core identity, precisely the same, 00:04:01.310 --> 00:04:03.129 even when you change the outfit, the background, 00:04:03.270 --> 00:04:05.409 the lighting, maybe even the expression slightly. 00:04:05.930 --> 00:04:07.669 This is the big problem with other tools like, 00:04:07.669 --> 00:04:11.729 say, DLA or Mid Journey. The face tends to wander 00:04:11.729 --> 00:04:14.219 with each new prompt. You know, I still wrestle 00:04:14.219 --> 00:04:16.620 with prompt drift myself sometimes, especially 00:04:16.620 --> 00:04:18.839 when I try to keep things consistent across different 00:04:18.839 --> 00:04:21.459 scenes, adding more details about the background 00:04:21.459 --> 00:04:24.879 or whatever. That subtle face change, it's tough 00:04:24.879 --> 00:04:26.660 to nail down. Yeah, it's a real frustration. 00:04:26.879 --> 00:04:28.800 And it just shows why starting simple is usually 00:04:28.800 --> 00:04:32.060 better. You upload a base image, a good clear 00:04:32.060 --> 00:04:35.019 headshot works best. That acts as the identity 00:04:35.019 --> 00:04:36.939 lock. Then you just start making variations, 00:04:37.259 --> 00:04:39.209 unlimited variations, really. Can we run through 00:04:39.209 --> 00:04:41.389 a couple of examples? Like, how different can 00:04:41.389 --> 00:04:43.649 the variations be using the same face? For sure. 00:04:43.769 --> 00:04:45.810 Take a simple headshot. Example one, you prompt. 00:04:46.389 --> 00:04:48.529 Change the shirt to a dark red turtleneck sweater. 00:04:48.990 --> 00:04:51.189 Put him in a cozy coffee shop with dim lights. 00:04:51.310 --> 00:04:54.649 Boom. Same face. Totally new context. Takes like 00:04:54.649 --> 00:04:57.449 15 seconds. And what if I need that same person, 00:04:57.529 --> 00:04:59.470 that same face, but for like a professional pitch 00:04:59.470 --> 00:05:03.110 video? Easy. Same face, new prompt. Change the 00:05:03.110 --> 00:05:05.470 outfit to a sharp gray suit and tie. The background 00:05:05.470 --> 00:05:08.079 is a modern office skyscraper lobby. Bright window 00:05:08.079 --> 00:05:11.500 light, done. You could honestly generate 50 versions 00:05:11.500 --> 00:05:14.519 for 50 different uses in maybe an hour. The source 00:05:14.519 --> 00:05:16.120 has also mentioned trying to avoid that kind 00:05:16.120 --> 00:05:19.079 of generic, too polished AI look. What are the 00:05:19.079 --> 00:05:21.120 best practices for prompts to get more realism 00:05:21.120 --> 00:05:23.800 with Nano Banana? OK, yeah, good point. Be specific 00:05:23.800 --> 00:05:25.740 with the details. Like, don't just say jacket, 00:05:25.920 --> 00:05:28.910 say black leather jacket. And. Crucially, you 00:05:28.910 --> 00:05:31.970 have to proactively add phrases about realism. 00:05:32.050 --> 00:05:35.389 Things like natural imperfections or subtle skin 00:05:35.389 --> 00:05:37.790 texture. Oh, OK. And definitely avoid telling 00:05:37.790 --> 00:05:40.769 the AI to make it flawless or perfect. That's 00:05:40.769 --> 00:05:42.970 like a direct ticket to the uncanny valley. It 00:05:42.970 --> 00:05:44.850 looks weird. Right. So if you had to pick just 00:05:44.850 --> 00:05:46.769 one prompt phrase, what's the most important 00:05:46.769 --> 00:05:50.329 one for fighting that super smooth, uncanny AI 00:05:50.329 --> 00:05:53.370 look? Adding subtle texture details like film 00:05:53.370 --> 00:05:56.410 grain combats the uncanny to perfect AI look. 00:05:56.670 --> 00:06:00.199 Film grain. interesting okay moving to step two 00:06:00.199 --> 00:06:02.600 Chad GPT we're using it for more than just writing 00:06:02.600 --> 00:06:04.459 the script right you said like the architect 00:06:04.459 --> 00:06:06.279 yeah exactly we use it to build the characters 00:06:06.279 --> 00:06:10.000 do it okay scaffolding let's call it like what's 00:06:10.000 --> 00:06:12.439 their persona is this avatar a serious finance 00:06:12.439 --> 00:06:15.279 analyst or I don't know it chill fitness guru. 00:06:15.699 --> 00:06:17.680 ChatGPT helps define their style, maybe common 00:06:17.680 --> 00:06:20.079 phrases they'd use. It makes them feel more believable 00:06:20.079 --> 00:06:22.579 than just some random face. And it also helps 00:06:22.579 --> 00:06:25.459 engineer the prompts for Nano Banana. Massively 00:06:25.459 --> 00:06:27.899 speeds things up. You can tell ChatGPT, hey, 00:06:27.899 --> 00:06:30.459 I need five detailed prompt ideas for professional 00:06:30.459 --> 00:06:32.800 tennis coach avatar. Give me different settings. 00:06:33.000 --> 00:06:35.579 And boom, it structures them for you. On the 00:06:35.579 --> 00:06:37.459 court, in the locker room, doing a practice drill, 00:06:37.540 --> 00:06:39.420 maybe an interview setup, you're not staring 00:06:39.420 --> 00:06:41.079 at a blank page trying to think of variations 00:06:41.079 --> 00:06:43.779 every single time. This is also where we tackle 00:06:43.779 --> 00:06:46.740 that robot voice problem you hear in so many 00:06:46.740 --> 00:06:49.660 AI videos. Why do they often sound so stiff? 00:06:50.279 --> 00:06:52.120 Well, usually it's because people write the scripts 00:06:52.120 --> 00:06:55.180 like they're writing a formal email. Yeah. All 00:06:55.180 --> 00:06:57.680 proper and buttoned up. Right. We have to specifically 00:06:57.680 --> 00:07:01.699 tell chat GPT, use a casual tone, use the first 00:07:01.699 --> 00:07:04.639 person, make the avatar say I. and really aim 00:07:04.639 --> 00:07:07.399 for a conversational flow, like how people actually 00:07:07.399 --> 00:07:09.860 talk. And even if ChatGPT drafts the script, 00:07:10.399 --> 00:07:12.980 those editing tips you mentioned sound, well, 00:07:13.800 --> 00:07:15.920 non -negotiable. Oh, absolutely critical. Use 00:07:15.920 --> 00:07:18.959 simple words like use instead of utilize. Keep 00:07:18.959 --> 00:07:21.459 sentences short, lots of contractions, don't, 00:07:21.459 --> 00:07:23.839 can't. That's conversational. And the number 00:07:23.839 --> 00:07:25.819 one test, read the script out loud yourself. 00:07:26.220 --> 00:07:28.899 If you stumble over the words, or if it sounds 00:07:28.899 --> 00:07:31.199 formal coming out of your mouth, the AI voice 00:07:31.199 --> 00:07:33.329 is going to sound even more unnatural. So it's 00:07:33.329 --> 00:07:35.889 not just generating prompts, it's building this 00:07:35.889 --> 00:07:40.050 reusable character library framework. Beyond 00:07:40.050 --> 00:07:43.470 the scripting part, how else does ChatGPT really 00:07:43.470 --> 00:07:46.189 save time in this whole avatar creation process? 00:07:46.649 --> 00:07:49.329 It dramatically speeds up the process by creating 00:07:49.329 --> 00:07:53.209 reusable, customizable prompt templates immediately. 00:07:53.470 --> 00:07:56.180 Got it. Reusable templates, makes sense. Minroll 00:07:56.180 --> 00:07:58.540 sponsor, Read Placeholder. Welcome back to the 00:07:58.540 --> 00:08:00.639 Deep Dive. Okay, so we've got our consistent 00:08:00.639 --> 00:08:03.759 custom avatar images made with NanoBanana, and 00:08:03.759 --> 00:08:06.379 we've planned out a natural conversational script 00:08:06.379 --> 00:08:09.660 using Chad GPT's help. Now for the final step. 00:08:10.200 --> 00:08:12.879 actually bringing that avatar to life using VEO 00:08:12.879 --> 00:08:15.500 3 .1 inside Heygen. Right, this is where it all 00:08:15.500 --> 00:08:17.759 comes together. We take that nice clean avatar 00:08:17.759 --> 00:08:19.899 image we made for free, remember using Nano Banana, 00:08:20.120 --> 00:08:21.899 and we upload that into Heygen. We're effectively 00:08:21.899 --> 00:08:25.180 skipping Heygen's own more expensive avatar creation 00:08:25.180 --> 00:08:28.839 step entirely. And this integration of Google's 00:08:28.839 --> 00:08:32.399 VEO 3 .1 model into Heygen, it's being talked 00:08:32.399 --> 00:08:35.210 about as a real leap forward in realism. What 00:08:35.210 --> 00:08:38.850 specifically makes VEO 3 .1 technically better 00:08:38.850 --> 00:08:41.289 than the older models? It really comes down to 00:08:41.289 --> 00:08:44.250 about three key upgrades. First is just superior 00:08:44.250 --> 00:08:47.830 realism. It renders in sharp 1080p, yeah. But 00:08:47.830 --> 00:08:50.049 the big thing is it uses a much larger understanding 00:08:50.049 --> 00:08:52.470 of context to figure out lighting and textures. 00:08:53.029 --> 00:08:55.269 So you get these hyper -natural shadows, subtle 00:08:55.269 --> 00:08:58.490 skin details. It looks much more lifelike. Okay, 00:08:58.769 --> 00:09:01.159 better textures and light. What's second? Second 00:09:01.159 --> 00:09:03.460 is Perfect Sync. The audio track and the video, 00:09:03.519 --> 00:09:05.580 the lip movements, they're generated together 00:09:05.580 --> 00:09:07.840 as one unit. This pretty much eliminates that 00:09:07.840 --> 00:09:09.980 weird robotic mouth movement or the slight delay 00:09:09.980 --> 00:09:11.799 you used to see all the time. Yeah, that lip 00:09:11.799 --> 00:09:13.899 sync issue was always the dead giveaway you were 00:09:13.899 --> 00:09:16.259 watching an AI video. It totally was. And the 00:09:16.259 --> 00:09:18.039 third thing is branding consistency. Because 00:09:18.039 --> 00:09:20.639 we're starting with that locked nano -bena image, 00:09:21.100 --> 00:09:23.340 VDO 3 .1 does a much better job of maintaining 00:09:23.340 --> 00:09:25.419 that exact look across lots of separate videos, 00:09:25.559 --> 00:09:27.759 which is huge if you're doing, say, a marketing 00:09:27.759 --> 00:09:29.919 campaign and need the same spokesperson intended. 00:09:29.929 --> 00:09:32.850 different ads. That claim, perfect sync, sounds 00:09:32.850 --> 00:09:35.789 almost too good to be true. Are there edge cases 00:09:35.789 --> 00:09:39.190 where VEO 3 .1 might still have trouble, like 00:09:39.190 --> 00:09:41.730 really fast talking or strong accents? It's definitely 00:09:41.730 --> 00:09:44.909 excellent VEO 3 .1 is. But yeah, extremely rapid 00:09:44.909 --> 00:09:47.210 speech or maybe some very distinct localized 00:09:47.210 --> 00:09:49.730 accents, those can still occasionally challenge 00:09:49.730 --> 00:09:52.049 the model a little bit. The safest approach is 00:09:52.049 --> 00:09:54.750 just to keep the scripts at a pretty normal conversational 00:09:54.750 --> 00:09:57.730 speaking pace. Whoa. But just imagine scaling 00:09:57.730 --> 00:10:01.009 this whole system. You could potentially produce, 00:10:01.009 --> 00:10:03.669 I don't know, thousands of localized video ads 00:10:03.669 --> 00:10:05.730 really quickly, like for global campaigns. You 00:10:05.730 --> 00:10:08.169 could A -B test different emotional deliveries, 00:10:08.389 --> 00:10:10.830 different regional languages, all while using 00:10:10.830 --> 00:10:13.350 the exact same spokesperson's face. That consistency, 00:10:13.450 --> 00:10:16.159 that's like... Exponential value. It completely 00:10:16.159 --> 00:10:18.340 changes the ROI calculation for your creative 00:10:18.340 --> 00:10:20.440 assets, yeah. And because we're only paying Hagen 00:10:20.440 --> 00:10:22.659 for that final render step, we're skipping the 00:10:22.659 --> 00:10:24.799 most credit -intensive part of the process. OK, 00:10:24.820 --> 00:10:27.700 on a practical note, what's the longest continuous 00:10:27.700 --> 00:10:30.840 video clip VEO 3 .1 can currently generate in 00:10:30.840 --> 00:10:32.960 Hagen before you have to cut and start a new 00:10:32.960 --> 00:10:37.120 segment? VEO 3 .1 allows for single continuous 00:10:37.120 --> 00:10:40.340 video segments up to 60 seconds long. 60 seconds. 00:10:40.379 --> 00:10:43.940 Oh, no. OK. Good to know. Now let's talk troubleshooting. 00:10:44.120 --> 00:10:47.179 Because when you set up a system like this, multi 00:10:47.179 --> 00:10:50.139 -tool systems, things inevitably go wrong sometimes. 00:10:50.179 --> 00:10:53.000 People hit snacks. Instead of just listing problems, 00:10:53.120 --> 00:10:55.860 maybe we can group the fixes. First, what about 00:10:55.860 --> 00:10:57.879 fixing issues with the image quality coming out 00:10:57.879 --> 00:10:59.860 of NanoBanana? OK, yeah. If NanoBanana gives 00:10:59.860 --> 00:11:02.519 you a weird face, like distorted features or 00:11:02.519 --> 00:11:05.159 something odd, the first fix is usually just 00:11:05.159 --> 00:11:07.340 simplify your prompt. Don't try to describe the 00:11:07.340 --> 00:11:09.460 entire universe in one go. Right. Focus on the 00:11:09.460 --> 00:11:12.029 face first. A young woman. Neutral expression, 00:11:12.309 --> 00:11:14.509 soft lighting, get that right. Then start adding 00:11:14.509 --> 00:11:17.009 complexity like outfits or backgrounds. And what 00:11:17.009 --> 00:11:19.309 if you get a great image from Nano Banana, you're 00:11:19.309 --> 00:11:21.590 happy with it, but then Hagen rejects it when 00:11:21.590 --> 00:11:24.009 you try to upload. That's almost always resolution. 00:11:24.289 --> 00:11:26.590 Hagen needs a minimum image size, typically a 00:11:26.590 --> 00:11:30.269 24 by 124 pixels. So the fix is simple. Before 00:11:30.269 --> 00:11:32.610 you upload to Hagen, run your Nano Banana image 00:11:32.610 --> 00:11:35.700 through a free online upscaler tool. OK, good 00:11:35.700 --> 00:11:37.580 tip. It just blows the image up to the right 00:11:37.580 --> 00:11:39.820 size. Don't skip that prep step. It saves headaches. 00:11:40.139 --> 00:11:41.580 All right, now for some of the more advanced 00:11:41.580 --> 00:11:43.500 stuff, the pro tips from the source material, 00:11:44.000 --> 00:11:46.460 let's talk about building an avatar team. What's 00:11:46.460 --> 00:11:49.740 the idea there? Yeah, this is cool for maybe 00:11:49.740 --> 00:11:51.840 an explainer video series where you want multiple 00:11:51.840 --> 00:11:54.860 hosts that look like they belong together. The 00:11:54.860 --> 00:11:57.919 key is what the source called. prompt parallelism 00:11:57.919 --> 00:12:00.580 prompt parallelism yeah basically you create 00:12:00.580 --> 00:12:02.259 the different characters but you make sure the 00:12:02.259 --> 00:12:04.919 prompts you use are structurally very similar 00:12:04.919 --> 00:12:07.440 like use the same lighting style description 00:12:07.440 --> 00:12:10.340 the same overall camera angle phrasing you only 00:12:10.340 --> 00:12:13.139 change small details hair color shirt color maybe 00:12:13.139 --> 00:12:15.659 slight facial structure tweaks they end up looking 00:12:15.659 --> 00:12:18.879 like they're from the same show or universe ah 00:12:18.879 --> 00:12:21.039 maintaining stylistic consistency through the 00:12:21.039 --> 00:12:24.059 prompts makes sense you also mentioned combining 00:12:24.059 --> 00:12:27.190 image elements like taking a hairstyle from one 00:12:27.190 --> 00:12:29.190 photo and an outfit from another. Yeah, that's 00:12:29.190 --> 00:12:31.549 leveraging Nano Banana's ability to use reference 00:12:31.549 --> 00:12:33.730 images. You can point it to one picture and say, 00:12:33.929 --> 00:12:35.970 use this hairstyle, point to another and say, 00:12:36.049 --> 00:12:38.230 use this jacket, all while keeping your main 00:12:38.230 --> 00:12:41.429 avatar's face locked from the original upload. 00:12:42.129 --> 00:12:45.210 It's pretty powerful for customization. And the 00:12:45.210 --> 00:12:49.529 last practical fix, dealing with small errors. 00:12:49.990 --> 00:12:52.769 Like if the AI generates weird fingers or strange 00:12:52.769 --> 00:12:55.659 eyes, but the rest of the image is perfect. Yeah, 00:12:55.659 --> 00:12:57.440 that's where inpainting comes in. Inpainting 00:12:57.440 --> 00:13:00.500 basically means you tell the AI, hey, regenerate 00:13:00.500 --> 00:13:02.600 only this tiny spot right here, like the weird 00:13:02.600 --> 00:13:04.759 finger or shadow that looks off without messing 00:13:04.759 --> 00:13:06.820 up the rest of the picture. Tools like Canva 00:13:06.820 --> 00:13:09.120 or the Photoshop Beta have features for this. 00:13:09.299 --> 00:13:11.139 It saves you re -rendering the whole thing just 00:13:11.139 --> 00:13:13.500 to fix one little glitch. OK, so when you're 00:13:13.500 --> 00:13:15.840 building that avatar team and aiming for a cohesive 00:13:15.840 --> 00:13:18.159 look, what's the single most important element 00:13:18.159 --> 00:13:20.240 to keep similar across the prompts for the different 00:13:20.240 --> 00:13:22.639 characters? Keeping the prompts similar ensures 00:13:22.639 --> 00:13:25.059 a matching look across the entire cast of avatars. 00:13:25.240 --> 00:13:27.500 Right, the stylistic elements in the prompt. 00:13:27.919 --> 00:13:30.720 So let's zoom out. What does this whole three 00:13:30.720 --> 00:13:33.659 -tool blueprint ultimately give someone, a content 00:13:33.659 --> 00:13:36.840 creator, a marketer? What's the big, so what? 00:13:37.059 --> 00:13:38.779 Well, fundamentally, it delivers efficiency, 00:13:38.899 --> 00:13:41.159 quality, and adaptability. You're getting genuine 00:13:41.159 --> 00:13:43.799 cost savings. You're getting effectively unlimited 00:13:43.799 --> 00:13:46.340 creative variations on your avatars. And you're 00:13:46.340 --> 00:13:48.519 getting really high visual quality, thanks to 00:13:48.519 --> 00:13:52.279 that VEO 3 .1 integration. Practically, it means 00:13:52.279 --> 00:13:54.539 you can easily A -B test different avatar styles. 00:13:54.730 --> 00:13:57.889 maybe male versus female, casual versus formal. 00:13:57.990 --> 00:14:00.090 You can do that across different campaigns, different 00:14:00.090 --> 00:14:02.049 regions, different languages, even. So it shifts 00:14:02.049 --> 00:14:05.110 the avatar from just being a visual in one video 00:14:05.110 --> 00:14:07.250 to being more like a strategic asset you can 00:14:07.250 --> 00:14:09.649 deploy in many ways. For creators, maybe they 00:14:09.649 --> 00:14:11.669 can build a consistent brand with multiple hosts 00:14:11.669 --> 00:14:13.850 without needing actors or complex licensing. 00:14:14.070 --> 00:14:15.730 Exactly. It's like building an asset library 00:14:15.730 --> 00:14:18.350 powerhouse. It really transforms your whole approach. 00:14:18.370 --> 00:14:21.710 You go from slow, expensive, and limited to being 00:14:21.710 --> 00:14:25.710 rapid, strategic, and pretty expansive in what 00:14:25.710 --> 00:14:28.070 you can create. You know, the real takeaway for 00:14:28.070 --> 00:14:30.070 me here isn't just the cost savings, though that's 00:14:30.070 --> 00:14:33.350 huge. It's the speed, the potential speed of 00:14:33.350 --> 00:14:36.309 asset creation. The sources suggest once you 00:14:36.309 --> 00:14:38.289 get comfortable with this flow, you could generate 00:14:38.289 --> 00:14:42.500 maybe 20 or more unique ready to go avatar images 00:14:42.500 --> 00:14:45.559 in less than half an hour. Yeah, it totally shifts 00:14:45.559 --> 00:14:47.679 your focus. You stop worrying about budgeting 00:14:47.679 --> 00:14:49.899 credits minute by minute, and you start thinking 00:14:49.899 --> 00:14:52.299 about building this digital factory of potential 00:14:52.299 --> 00:14:56.340 spokespeople. Just imagine having, say, 50 unique 00:14:56.919 --> 00:14:59.679 Consistent characters ready to deploy. Tailored 00:14:59.679 --> 00:15:02.639 for any situation you need a serious lawyer avatar. 00:15:02.940 --> 00:15:05.559 Got it. An enthusiastic team for a different 00:15:05.559 --> 00:15:08.919 campaign. Got it. Fitness guru, corporate CEO, 00:15:09.460 --> 00:15:11.500 all sitting in your library ready to be turned 00:15:11.500 --> 00:15:14.320 into high quality video for fractions of what 00:15:14.320 --> 00:15:16.100 it used to cost. You could actually start building 00:15:16.100 --> 00:15:17.919 that library today. The tools are out there. 00:15:18.039 --> 00:15:20.379 They're mostly accessible. The method seems pretty 00:15:20.379 --> 00:15:22.360 well proven according to the source. Time to 00:15:22.360 --> 00:15:24.700 start creating. Out T -Row music.