WEBVTT 00:00:00.000 --> 00:00:02.299 Imagine stepping into the spotlight, starring 00:00:02.299 --> 00:00:05.339 in your own professional video ads, maybe creating 00:00:05.339 --> 00:00:07.700 compelling content for your brand or, you know, 00:00:07.820 --> 00:00:10.080 simply sharing a story with the world and doing 00:00:10.080 --> 00:00:12.380 all of that without ever having to spend hours 00:00:12.380 --> 00:00:14.539 and hours in front of a camera. This isn't science 00:00:14.539 --> 00:00:16.100 fiction anymore. It's actually happening right 00:00:16.100 --> 00:00:20.379 now. Welcome to the deep dive. Today, we're plunging 00:00:20.379 --> 00:00:23.600 into a true game changer. how artificial intelligence 00:00:23.600 --> 00:00:26.640 is making personal video production, while unbelievably 00:00:26.640 --> 00:00:29.539 accessible. We're talking about groundbreaking 00:00:29.539 --> 00:00:32.740 tools like Google VEO3, this fascinating concept 00:00:32.740 --> 00:00:35.920 of your very own digital twin, and how these 00:00:35.920 --> 00:00:38.500 things unlock a completely new world of creative 00:00:38.500 --> 00:00:40.840 possibility for you. We'll guide you through 00:00:40.840 --> 00:00:43.399 preparing your digital self, crafting those perfect 00:00:43.399 --> 00:00:45.520 prompts to bring your vision to life, breathing 00:00:45.520 --> 00:00:48.520 dynamic motion into still images, and finally, 00:00:48.799 --> 00:00:51.320 showing you how to tell a full, compelling narrative 00:00:51.320 --> 00:00:53.740 story. with your AI counterpart. Okay, let's 00:00:53.740 --> 00:00:56.479 untack this a bit. We are genuinely living through 00:00:56.479 --> 00:01:00.619 a breakthrough era where AI isn't some distant 00:01:00.619 --> 00:01:02.840 theoretical thing. It's becoming a really powerful 00:01:02.840 --> 00:01:05.359 creative tool. It's truly right at our fingertips. 00:01:05.920 --> 00:01:08.620 The ability to create incredibly realistic videos 00:01:08.620 --> 00:01:11.400 of yourself talking, presenting, sharing a story. 00:01:11.760 --> 00:01:15.019 It's gone from being a futuristic dream to, well, 00:01:15.159 --> 00:01:17.359 a practical reality. It's a pretty fundamental 00:01:17.359 --> 00:01:19.540 shift, I think. And what's truly fascinating 00:01:19.540 --> 00:01:22.540 here is the democratizing effect this technology 00:01:22.540 --> 00:01:24.620 has on video production. I mean, think about 00:01:24.620 --> 00:01:27.400 it. The traditional barriers, the sky -high costs 00:01:27.400 --> 00:01:29.719 of equipment, needing specialized technical skills, 00:01:30.200 --> 00:01:32.719 those endless hours of filming and editing, they're 00:01:32.719 --> 00:01:34.760 all starting to just dissolve. Instead of needing 00:01:34.760 --> 00:01:37.079 a huge budget or a full production crew, your 00:01:37.079 --> 00:01:39.620 most valuable assets now are really your creativity 00:01:39.620 --> 00:01:42.659 and, yeah, a few really high quality images of 00:01:42.659 --> 00:01:45.200 yourself. It's a remarkable change. It essentially 00:01:45.200 --> 00:01:47.489 empowers any individual to become their own mini 00:01:47.489 --> 00:01:50.980 studio. It's kind of wild. That's precisely what 00:01:50.980 --> 00:01:54.000 this deep dive is all about. We're going to explore 00:01:54.000 --> 00:01:58.620 the exact process of harnessing Google VEO3 alongside 00:01:58.620 --> 00:02:01.640 other powerful AI tools to create professional 00:02:01.640 --> 00:02:04.439 quality videos where you are the central character. 00:02:04.920 --> 00:02:07.019 Our focus isn't just on the how, you know, the 00:02:07.019 --> 00:02:10.219 steps, but also on the deeper why behind each 00:02:10.219 --> 00:02:12.319 crucial stage. We want to help you not just follow 00:02:12.319 --> 00:02:14.840 instructions, but to genuinely understand and 00:02:14.840 --> 00:02:18.030 master this revolutionary technology. It's a 00:02:18.030 --> 00:02:20.349 powerful moment, honestly, allowing a single 00:02:20.349 --> 00:02:22.669 individual to compete with entire production 00:02:22.669 --> 00:02:25.229 studios, blurring those lines between amateur 00:02:25.229 --> 00:02:27.430 and professional output in a way we've just never 00:02:27.430 --> 00:02:30.710 seen before. To begin creating this magic, the 00:02:30.710 --> 00:02:33.009 first and probably most crucial step is actually 00:02:33.009 --> 00:02:35.849 gaining access to Google VEO3's latest feature, 00:02:36.150 --> 00:02:38.729 its image -to -video conversion. This is the 00:02:38.729 --> 00:02:40.610 really revolutionary update that allows you to 00:02:40.610 --> 00:02:42.509 upload your own photos and then transform them 00:02:42.509 --> 00:02:44.830 into talking videos with pretty impressive lip 00:02:44.830 --> 00:02:46.689 -syncing capabilities. That level of realism 00:02:46.689 --> 00:02:48.870 just wasn't possible in earlier versions, not 00:02:48.870 --> 00:02:52.199 like this. You got it. Now currently, because 00:02:52.199 --> 00:02:55.539 this feature is in an early phase rollout and 00:02:55.539 --> 00:02:58.060 testing phase, there are some specific access 00:02:58.060 --> 00:03:00.180 requirements. For instance, you'll probably need 00:03:00.180 --> 00:03:03.960 a US -based email account, one created while 00:03:03.960 --> 00:03:06.259 you were physically in the United States. Two 00:03:06.259 --> 00:03:09.360 secs silence. Google tends to prioritize its 00:03:09.360 --> 00:03:11.680 domestic market for gathering initial feedback 00:03:11.680 --> 00:03:14.659 in a controlled environment before a global expansion. 00:03:15.379 --> 00:03:17.969 Makes sense. Also, you'll likely need a virtual 00:03:17.969 --> 00:03:20.990 private network or VPN connected to a US server. 00:03:21.050 --> 00:03:23.509 This is simply to bypass those regional restrictions. 00:03:23.889 --> 00:03:25.550 It makes the system believe you're accessing 00:03:25.550 --> 00:03:28.229 it from within the US, which then unlocks that 00:03:28.229 --> 00:03:30.090 image upload feature. Without that, you'd probably 00:03:30.090 --> 00:03:32.490 just run into error messages or not see the option 00:03:32.490 --> 00:03:34.789 at all. So these regional restrictions are just 00:03:34.789 --> 00:03:37.050 a temporary gateway. Yeah, exactly. Therefore, 00:03:37.330 --> 00:03:40.090 controlled testing before what we expect will 00:03:40.090 --> 00:03:42.750 be a global rollout eventually. OK, here's where 00:03:42.750 --> 00:03:45.389 it gets really interesting. And frankly, it's 00:03:45.389 --> 00:03:47.509 arguably the most critical foundational stage 00:03:47.509 --> 00:03:50.110 in the entire process, building your digital 00:03:50.110 --> 00:03:54.030 twin. The quality of your final AI video is almost 00:03:54.030 --> 00:03:56.590 entirely dependent on the quality and consistency 00:03:56.590 --> 00:03:58.870 of your source images. We're not just taking 00:03:58.870 --> 00:04:00.870 photos here. We're essentially building an AI 00:04:00.870 --> 00:04:02.710 version of yourself that can then be directed 00:04:02.710 --> 00:04:04.710 and placed into any context you can imagine. 00:04:05.479 --> 00:04:08.479 A well -trained digital twin allows the AI to 00:04:08.479 --> 00:04:10.900 replicate you with incredible fidelity, capturing 00:04:10.900 --> 00:04:13.060 your facial features, your expressions, even 00:04:13.060 --> 00:04:15.539 those smallest characteristic details. So what 00:04:15.539 --> 00:04:17.379 are the golden rules for shooting these reference 00:04:17.379 --> 00:04:20.500 photos? Right. First, variety is key. The AI 00:04:20.500 --> 00:04:22.660 learns kind of like a human, but at warp speed. 00:04:22.980 --> 00:04:25.000 So the more diverse examples of you it sees, 00:04:25.120 --> 00:04:27.240 the better it understands who you truly are. 00:04:27.540 --> 00:04:30.860 Second, focus on angles. Shoot straight on, three 00:04:30.860 --> 00:04:32.860 quarters views from both left and right, full 00:04:32.860 --> 00:04:35.100 profile, and then slightly from above. and slightly 00:04:35.100 --> 00:04:37.480 from below. This comprehensive approach helps 00:04:37.480 --> 00:04:39.819 the AI capture your unique facial structure. 00:04:40.120 --> 00:04:42.519 It needs to see you from all sides, basically. 00:04:43.759 --> 00:04:45.860 Third, and this is absolutely crucial for giving 00:04:45.860 --> 00:04:48.060 your final video that, you know, soul, you need 00:04:48.060 --> 00:04:51.040 a full range of expressions. Think happy, sad, 00:04:51.319 --> 00:04:53.639 surprised, thoughtful, a small smile, a wide 00:04:53.639 --> 00:04:56.199 laugh. Don't just give it a neutral face. The 00:04:56.199 --> 00:04:58.319 AI needs to see your emotional range to replicate 00:04:58.319 --> 00:05:01.740 it. Fourth, consider lighting. Take photos in 00:05:01.740 --> 00:05:03.879 various conditions, soft natural light near a 00:05:03.879 --> 00:05:06.600 window, maybe some harsh outdoor light, warm 00:05:06.600 --> 00:05:09.839 indoor light. This helps the AI render you realistically 00:05:09.839 --> 00:05:12.240 in different scenarios, making your twin highly 00:05:12.240 --> 00:05:15.180 adaptable. Fifth, it's really about quality over 00:05:15.180 --> 00:05:18.060 quantity. 20 sharp, clear, noise -free images 00:05:18.060 --> 00:05:20.680 are far, far better than 100 blurry or poorly 00:05:20.680 --> 00:05:23.180 lit ones. Make sure your face is in focus and 00:05:23.180 --> 00:05:24.839 takes up a significant portion of the frame. 00:05:25.500 --> 00:05:27.100 And finally, for that initial training, maybe 00:05:27.100 --> 00:05:29.680 wear a neutral wardrobe. simple solid colors 00:05:29.680 --> 00:05:32.139 like gray, black, or white. This helps the AI 00:05:32.139 --> 00:05:33.980 focus on learning your facial features without 00:05:33.980 --> 00:05:36.160 being distracted by complex patterns. You can 00:05:36.160 --> 00:05:37.819 always change the clothing later with trumps, 00:05:38.279 --> 00:05:40.420 but getting the face right first is really vital. 00:05:40.660 --> 00:05:43.220 Yeah, totally. And to help with this, one excellent 00:05:43.220 --> 00:05:46.439 tool is Higgs Field AI. This is ideal if you 00:05:46.439 --> 00:05:48.779 prioritize speed, simplicity, and crucially, 00:05:49.199 --> 00:05:52.040 character consistency across multiple image generations. 00:05:52.540 --> 00:05:54.860 The process involves preparing a deliberate training 00:05:54.860 --> 00:05:57.420 data set. We kind of recommend a starter pack 00:05:57.420 --> 00:06:01.079 of about 25 photos. That includes like a 360 00:06:01.079 --> 00:06:03.579 degree portrait set where you keep the camera 00:06:03.579 --> 00:06:05.740 still and slowly turn your head, giving maybe 00:06:05.740 --> 00:06:08.939 nine distinct angles. Then an expression gauntlet, 00:06:09.019 --> 00:06:10.879 which is maybe eight photos of basic emotions 00:06:10.879 --> 00:06:14.220 like joy, sadness, surprise, a natural smile, 00:06:14.259 --> 00:06:16.319 things like that. And a wardrobe collection, 00:06:16.480 --> 00:06:18.160 say eight medium shots and two or three different 00:06:18.160 --> 00:06:20.759 outfits you might typically use. You upload these 00:06:20.759 --> 00:06:23.319 to Higgs Field. The system analyzes them, gives 00:06:23.319 --> 00:06:26.399 you a quality score. slow you simply review and 00:06:26.399 --> 00:06:29.040 replace those poor images. pretty straightforward. 00:06:29.420 --> 00:06:31.199 Once your character is created, you test and 00:06:31.199 --> 00:06:32.779 refine it with simple prompts, adding more photos 00:06:32.779 --> 00:06:35.279 up to 70, I think, if needed to improve the accuracy. 00:06:35.540 --> 00:06:37.519 It's a quick and powerful way to get started. 00:06:37.819 --> 00:06:39.519 Then there's mid -journey. This is maybe for 00:06:39.519 --> 00:06:42.920 those seeking a more photographic or artistic 00:06:42.920 --> 00:06:45.519 -level image quality. This method embraces a 00:06:45.519 --> 00:06:48.100 bit more trial and error to achieve that perfect 00:06:48.100 --> 00:06:50.579 consistency. The key here is mastering how to 00:06:50.579 --> 00:06:53.040 combine your image reference with really detailed 00:06:53.040 --> 00:06:55.879 text prompts. You provide an image URL, then 00:06:55.879 --> 00:06:58.350 describe exactly what you want. A professional 00:06:58.350 --> 00:07:00.629 headshot of a person in their early 30s wearing 00:07:00.629 --> 00:07:03.569 a dark gray merino wool turtleneck. Maybe adding 00:07:03.569 --> 00:07:06.089 details about the setting, like a modern minimalist 00:07:06.089 --> 00:07:08.910 office with soft diffused window light. You can 00:07:08.910 --> 00:07:11.529 even specify cinematic qualities, like telling 00:07:11.529 --> 00:07:15.310 the AI you want it shot with an 85mm f1 .8 lens, 00:07:15.670 --> 00:07:17.230 which creates that beautiful blurred background, 00:07:17.610 --> 00:07:20.550 or demand things like hyper -detailed, photorealistic, 00:07:20.670 --> 00:07:23.509 sharp focus on the eyes. This level of control 00:07:23.509 --> 00:07:25.970 helps achieve truly photographic quality while 00:07:25.970 --> 00:07:28.790 keeping your digital twin consistent. The precise 00:07:28.790 --> 00:07:30.670 parameters for things like character reference, 00:07:31.110 --> 00:07:33.709 cref, and weighting, CW, well, we'll put those 00:07:33.709 --> 00:07:34.990 details in our show notes for anyone who wants 00:07:34.990 --> 00:07:36.829 to dive really deep into the specific syntax. 00:07:37.089 --> 00:07:39.149 It gets pretty technical. And for the ultimate 00:07:39.149 --> 00:07:42.310 control, especially for professional users or 00:07:42.310 --> 00:07:44.509 those building a massive library of content, 00:07:45.209 --> 00:07:47.930 there's stable diffusion with LoRa. Okay, so 00:07:47.930 --> 00:07:49.769 Allora, that stands for low -rank adaptation. 00:07:50.250 --> 00:07:53.149 It's basically a small plug -in file. Think of 00:07:53.149 --> 00:07:55.410 it as a highly compressed file containing all 00:07:55.410 --> 00:07:58.089 the intricate data just about your face. Once 00:07:58.089 --> 00:08:00.410 you train it, which typically requires maybe 00:08:00.410 --> 00:08:03.550 20 to 30 images and a more technical process, 00:08:03.589 --> 00:08:06.029 got to admit, you can attach this Allora file 00:08:06.029 --> 00:08:08.610 to pretty much any stable diffusion model. This 00:08:08.610 --> 00:08:10.750 lets you generate images of yourself with near 00:08:10.750 --> 00:08:13.850 -perfect accuracy and consistency across countless 00:08:13.850 --> 00:08:16.310 scenarios. You'd use this when you need to create 00:08:16.310 --> 00:08:18.310 like hundreds of images of the same character 00:08:18.310 --> 00:08:20.550 in totally different scenes, outfits, and styles. 00:08:20.930 --> 00:08:23.110 It's the most labor -intensive method to set 00:08:23.110 --> 00:08:25.389 up initially, for sure, but it gives the highest 00:08:25.389 --> 00:08:27.750 long -term rewards for consistency and creative 00:08:27.750 --> 00:08:30.949 freedom. Whoa. I mean, imagine creating hundreds 00:08:30.949 --> 00:08:33.389 of perfectly consistent character images in totally 00:08:33.389 --> 00:08:35.509 different scenarios. That's kind of mind -bending 00:08:35.509 --> 00:08:37.529 when you think about it. So it really does come 00:08:37.529 --> 00:08:39.370 back to this. Getting those initial source images 00:08:39.370 --> 00:08:42.110 right is truly the bedrock. Yes, absolutely. 00:08:42.350 --> 00:08:44.970 High -quality images are vital for a realistic 00:08:44.970 --> 00:08:47.409 digital twin. and they are the absolute foundation, 00:08:47.509 --> 00:08:50.389 no shortcuts there. Moving on, writing a prompt 00:08:50.389 --> 00:08:52.710 isn't just about giving an order to an AI anymore. 00:08:53.129 --> 00:08:56.029 It's rapidly evolving into, well, a new art form, 00:08:56.210 --> 00:08:59.029 a new type of creative direction. Instead of 00:08:59.029 --> 00:09:01.350 struggling yourself to find the perfect technical 00:09:01.350 --> 00:09:04.070 terms for your image generation, why not use 00:09:04.070 --> 00:09:06.750 a language AI to actually direct the image AI? 00:09:07.269 --> 00:09:09.950 This is where models like ChatGBT or Google Gemini 00:09:09.950 --> 00:09:12.399 become your personal prompt engineer. It's almost 00:09:12.399 --> 00:09:14.539 like learning to speak to a creative collaborator 00:09:14.539 --> 00:09:16.980 in a new hyper -specific language, and it opens 00:09:16.980 --> 00:09:18.679 up possibilities you might not have even been 00:09:18.679 --> 00:09:20.840 able to articulate before. Yeah, this workflow 00:09:20.840 --> 00:09:23.679 is incredibly powerful for a few key reasons. 00:09:24.100 --> 00:09:27.779 First, specialized knowledge. AIs like ChatGPT 00:09:27.779 --> 00:09:30.100 are trained on these massive data sets that include 00:09:30.100 --> 00:09:33.919 comprehensive terminology from photography, cinematography, 00:09:34.080 --> 00:09:36.799 lighting, you name it. They know what an anamorphic 00:09:36.799 --> 00:09:38.960 lens does or what Rembrandt lighting looks like, 00:09:39.179 --> 00:09:40.980 way beyond what most of us have just memorized, 00:09:41.000 --> 00:09:43.799 right? Second, it helps in overcoming creative 00:09:43.799 --> 00:09:46.320 blocks. Sometimes you have a vision but just 00:09:46.320 --> 00:09:48.080 don't know how to describe it technically. You 00:09:48.080 --> 00:09:50.700 can simply describe it naturally and the AI will 00:09:50.700 --> 00:09:52.759 translate it into a detailed technical brief 00:09:52.759 --> 00:09:55.750 for the image generator. And third, linguistic 00:09:55.750 --> 00:09:58.470 optimization. The AI can generate prompts with 00:09:58.470 --> 00:10:00.830 a richer vocabulary and maybe more complex grammatical 00:10:00.830 --> 00:10:02.710 structures than we might typically think of, 00:10:02.870 --> 00:10:05.009 allowing you to really extract the maximum potential 00:10:05.009 --> 00:10:07.570 from the image model. The workflow itself approaches 00:10:07.570 --> 00:10:09.870 this process like a conversation with a creative 00:10:09.870 --> 00:10:12.809 expert, which is pretty cool. First, you prime 00:10:12.809 --> 00:10:15.549 the AI by establishing its role. You might say 00:10:15.549 --> 00:10:17.649 something like, you will act as a professional 00:10:17.649 --> 00:10:20.690 prompt engineer, an expert in creating detailed 00:10:20.690 --> 00:10:24.370 text prompts for generative image AIs. Your task 00:10:24.370 --> 00:10:27.049 is to transform my simple ideas into complex, 00:10:27.309 --> 00:10:29.889 visually rich, and technically optimized prompts. 00:10:30.389 --> 00:10:32.750 Always prioritize a cinematic style, dramatic 00:10:32.750 --> 00:10:36.690 lighting, and photorealism. Are you ready? Then 00:10:36.690 --> 00:10:38.929 you draft your idea naturally, describing the 00:10:38.929 --> 00:10:40.970 scene you want in plain language, just as if 00:10:40.970 --> 00:10:43.309 you were talking to a human director. For example, 00:10:43.830 --> 00:10:46.149 I want an image of myself using the character 00:10:46.149 --> 00:10:48.620 I've created. I'm standing in a high -tech workshop 00:10:48.620 --> 00:10:51.299 at night. Around me are electronic devices and 00:10:51.299 --> 00:10:53.919 ethereal holographic displays. The main light 00:10:53.919 --> 00:10:55.960 comes from blue and purple neon strips on the 00:10:55.960 --> 00:10:58.220 walls. I'm holding a gently glowing microchip 00:10:58.220 --> 00:10:59.799 and looking at it with a focused, passionate 00:10:59.799 --> 00:11:02.379 expression. I want it to look cool and futuristic. 00:11:02.779 --> 00:11:04.519 Pretty straightforward. Finally, you request 00:11:04.519 --> 00:11:07.299 the AI to generate a prompt and variations. You'd 00:11:07.299 --> 00:11:09.539 say something like, based on that idea, generate 00:11:09.539 --> 00:11:12.059 a detail prompt for Mid -Journey V6 using a character 00:11:12.059 --> 00:11:14.440 reference. Add details about the camera lens, 00:11:14.799 --> 00:11:16.539 the quality of light, and the characters in motion. 00:11:17.109 --> 00:11:19.389 Then give me two more variations of this prompt, 00:11:19.649 --> 00:11:21.769 one that focuses on a wider angle to show the 00:11:21.769 --> 00:11:24.470 whole workshop, and one with a closer, more intimate 00:11:24.470 --> 00:11:27.669 angle on my face and the chip. The AI can then 00:11:27.669 --> 00:11:30.450 produce something incredibly detailed, like cinematic 00:11:30.450 --> 00:11:32.870 medium shot of a male creator in his futuristic 00:11:32.870 --> 00:11:35.840 workshop at night. surrounded by glowing holographic 00:11:35.840 --> 00:11:39.539 displays, holding a small, softly glowing microprocessor. 00:11:39.960 --> 00:11:42.080 Intense and passionate expression, illuminated 00:11:42.080 --> 00:11:44.740 by dramatic blue and purple neon, shot on an 00:11:44.740 --> 00:11:47.779 Arri Alexa camera with a 50mm anamorphic lens, 00:11:48.259 --> 00:11:50.419 volumetric lighting, hyper -realistic textures, 00:11:50.779 --> 00:11:52.960 cyberpunk aesthetic. And then it adds the technical 00:11:52.960 --> 00:11:55.519 parameters like aspect ratio and style. I still 00:11:55.519 --> 00:11:57.500 wrestle with prompt drift myself sometimes, getting 00:11:57.500 --> 00:11:59.360 the AI to consistently do what I envisioned. 00:11:59.820 --> 00:12:02.019 So having an AI as a prompt engineer feels like 00:12:02.019 --> 00:12:04.200 a true creative partner. So it's like having 00:12:04.200 --> 00:12:07.419 a co -director for your AI. Precisely. It translates 00:12:07.419 --> 00:12:11.059 your vision into the AI's specific language with 00:12:11.059 --> 00:12:14.340 remarkable precision. It's really helpful. Okay, 00:12:14.419 --> 00:12:17.419 next up, let's talk about refining things with 00:12:17.419 --> 00:12:20.259 FreePic Picasso's Flux Context feature. You can 00:12:20.259 --> 00:12:22.419 really think of this as your digital retouching 00:12:22.419 --> 00:12:24.799 artist. This tech is based on something called 00:12:24.799 --> 00:12:27.620 inpainting. So what's inpainting? It basically 00:12:27.620 --> 00:12:29.799 means you can literally paint over a region of 00:12:29.799 --> 00:12:32.399 an image and then command the AI to intelligently 00:12:32.399 --> 00:12:34.679 replace or add something new right into that 00:12:34.679 --> 00:12:36.919 painted area. It's incredibly precise when you 00:12:36.919 --> 00:12:40.340 get it right. The process itself is quite straightforward. 00:12:40.799 --> 00:12:43.889 You upload your base image. that nearly perfect 00:12:43.889 --> 00:12:46.269 picture you've already created. Then you upload 00:12:46.269 --> 00:12:48.409 a context image, which is the object you want 00:12:48.409 --> 00:12:50.690 to introduce into your scene, say, a separate 00:12:50.690 --> 00:12:53.509 product shot on a plain white background. Finally, 00:12:53.529 --> 00:12:56.009 and this is key, you write a directorial prompt. 00:12:56.549 --> 00:12:58.389 This is where the magic really happens, because 00:12:58.389 --> 00:13:00.450 your prompt needs to be incredibly specific to 00:13:00.450 --> 00:13:03.269 get good results. For example, if you have a 00:13:03.269 --> 00:13:05.309 shot of yourself in a cafe in a separate picture 00:13:05.309 --> 00:13:07.549 of some headphones, you wouldn't just prompt, 00:13:07.549 --> 00:13:10.990 add headphones. That's too vague. A much better, 00:13:11.009 --> 00:13:12.950 more effective prompt would be something like, 00:13:13.350 --> 00:13:15.250 The man at the cafe is now holding the white 00:13:15.250 --> 00:13:17.769 wireless earbuds case in his right hand. His 00:13:17.769 --> 00:13:19.470 fingers are wrapped naturally around the case 00:13:19.470 --> 00:13:21.870 as he presents it towards the camera. Match the 00:13:21.870 --> 00:13:25.159 lighting of the cafe. See? That level of detail 00:13:25.159 --> 00:13:28.039 helps make the integration seamless. Now, while 00:13:28.039 --> 00:13:30.399 Flux context is strong in its speed and being 00:13:30.399 --> 00:13:32.299 web -based, it's worth mentioning that tools 00:13:32.299 --> 00:13:34.600 like Adobe Photoshop's generative fill perform 00:13:34.600 --> 00:13:37.360 a similar function, maybe offering deeper integration 00:13:37.360 --> 00:13:39.259 into a professional workflow and potentially 00:13:39.259 --> 00:13:42.759 more detailed control. But Flux is fast. So this 00:13:42.759 --> 00:13:45.100 is where you fine -tune the scene, adding or 00:13:45.100 --> 00:13:48.379 changing elements. Yes, exactly. It's intelligent 00:13:48.379 --> 00:13:51.539 retouching, making precise changes quickly. Mid 00:13:51.539 --> 00:13:53.940 -roll sponsor read. All right, now we get to 00:13:53.940 --> 00:13:56.200 the really exciting part. This is when you sit 00:13:56.200 --> 00:13:59.059 squarely in the director's chair. You're now 00:13:59.059 --> 00:14:01.779 directing the AI not only on what action takes 00:14:01.779 --> 00:14:04.879 place, but also on camera movement and the overall 00:14:04.879 --> 00:14:07.980 soul of the scene. So this is where the still 00:14:07.980 --> 00:14:10.480 image truly becomes a living scene. Exactly. 00:14:10.759 --> 00:14:13.139 It's where your creative vision translates into 00:14:13.139 --> 00:14:15.799 motion. An effective video prompt really needs 00:14:15.799 --> 00:14:18.179 to cover four key elements. OK, what are they? 00:14:18.299 --> 00:14:20.620 First, character action. What does the character 00:14:20.620 --> 00:14:23.100 actually do? Like, the character raises the perfume 00:14:23.100 --> 00:14:26.440 bottle. Simple enough. Second, dialogue. What 00:14:26.440 --> 00:14:28.580 does the character say? Maybe adds and says, 00:14:28.720 --> 00:14:32.000 experience the new scent. Third, camera movement. 00:14:32.480 --> 00:14:34.679 What does the camera do during this? The camera 00:14:34.679 --> 00:14:38.059 slowly pushes in on the character's face. And 00:14:38.059 --> 00:14:41.120 fourth, environmental animation. What else is 00:14:41.120 --> 00:14:43.139 moving in the surroundings? Something like, as 00:14:43.139 --> 00:14:45.559 a gentle breeze rustles the crops in the background. 00:14:45.899 --> 00:14:48.320 Thinking about all four helps you construct a 00:14:48.320 --> 00:14:50.840 complete living scene, not just a static image 00:14:50.840 --> 00:14:53.960 that moves slightly. Gotcha. So the process involves 00:14:53.960 --> 00:14:56.720 accessing Google Flow and selecting frames to 00:14:56.720 --> 00:14:59.139 video. You upload your perfect starting still 00:14:59.139 --> 00:15:01.759 image, the one you painstakingly crafted earlier, 00:15:02.059 --> 00:15:04.200 then you construct that director's prompt, often 00:15:04.200 --> 00:15:06.519 using something like Chat GPT, to help combine 00:15:06.519 --> 00:15:08.200 all four of those elements we just discussed. 00:15:08.539 --> 00:15:11.200 Right. So for example, for maybe a five second 00:15:11.200 --> 00:15:13.940 tech review video, your prompt might be something 00:15:13.940 --> 00:15:17.080 like... Using this image as a starting point 00:15:17.080 --> 00:15:20.259 for Google VEO 3, create a five -second video. 00:15:20.779 --> 00:15:23.360 The character, a tech reviewer, should lift the 00:15:23.360 --> 00:15:25.460 silver smartwatch towards the camera, turning 00:15:25.460 --> 00:15:28.120 his wrist to catch the light. He should say with 00:15:28.120 --> 00:15:31.799 a confident smile, this changes everything. The 00:15:31.799 --> 00:15:34.419 camera will perform a slow, dolly zoom, starting 00:15:34.419 --> 00:15:36.519 as a medium shot and ending as a close -up on 00:15:36.519 --> 00:15:39.279 the watch and his face. In the background, the 00:15:39.279 --> 00:15:41.820 neon lights should have a subtle, pulsating glow. 00:15:42.059 --> 00:15:44.240 See how specific that is. Yeah, that covers everything. 00:15:44.539 --> 00:15:46.799 And you mentioned managing resources or credits 00:15:46.799 --> 00:15:49.299 within VE03. Oh yeah, important point. You'll 00:15:49.299 --> 00:15:52.059 probably want to use VE03 fast for drafts, for 00:15:52.059 --> 00:15:53.860 testing movements and ideas, almost like doing 00:15:53.860 --> 00:15:56.000 pre -visualization. It's cheaper on credits. 00:15:56.480 --> 00:15:59.799 Then reserve VE03 quality for your final high 00:15:59.799 --> 00:16:02.200 -detail shots, especially those close -ups where 00:16:02.200 --> 00:16:04.100 quality really matters. Okay, that makes sense. 00:16:04.379 --> 00:16:06.820 So a truly compelling video isn't just one clip, 00:16:06.960 --> 00:16:09.779 right? It's made from many individual shots arranged 00:16:09.779 --> 00:16:12.639 purposefully to tell a story. This step sounds 00:16:12.639 --> 00:16:15.539 like building a film piece by piece. It is, yeah. 00:16:15.700 --> 00:16:17.899 It's cinematic storytelling. Planning your shots 00:16:17.899 --> 00:16:20.340 before you shoot with the AI. Precisely. And 00:16:20.340 --> 00:16:22.460 this means shifting your thinking from just a 00:16:22.460 --> 00:16:25.039 simple shot list, which is merely a text list. 00:16:25.480 --> 00:16:28.159 Scene one, wide shot scene two, medium shot scene 00:16:28.159 --> 00:16:31.879 three. Close up to creating a storyboard. A storyboard 00:16:31.879 --> 00:16:34.340 is the visual version of that shot list, and 00:16:34.340 --> 00:16:35.879 you don't need to be an artist. Simple stick 00:16:35.879 --> 00:16:38.539 figures are perfectly fine, seriously. The goal 00:16:38.539 --> 00:16:40.980 is just to visualize the composition, the camera 00:16:40.980 --> 00:16:43.240 angle, the action, and the flow of each scene 00:16:43.240 --> 00:16:45.259 before you start generating the video clips. 00:16:45.639 --> 00:16:47.519 This helps ensure your story flows logically 00:16:47.519 --> 00:16:49.740 before you start spending those valuable VEO 00:16:49.740 --> 00:16:52.440 credits. Right. The process that involves creating 00:16:52.440 --> 00:16:55.500 each of those shots one by one using the image 00:16:55.500 --> 00:16:57.379 generation tools we talked about, Mid Journey, 00:16:57.539 --> 00:17:00.299 Higgs Field, whatever works best for you, ensuring 00:17:00.299 --> 00:17:02.659 absolute consistency in lighting, clothing, and 00:17:02.659 --> 00:17:05.079 setting across the shots. That consistency is 00:17:05.079 --> 00:17:07.900 key. Then you bring each of those perfectly consistent 00:17:07.900 --> 00:17:10.900 still images into Google VEO 3 to turn them into 00:17:10.900 --> 00:17:12.779 the short video clips that make up your narrative. 00:17:13.259 --> 00:17:15.319 It's a bit like stacking Lego blocks of data, 00:17:15.640 --> 00:17:17.640 building your film piece by piece, shot by shot. 00:17:17.900 --> 00:17:19.359 OK, so you've got your visuals. The clips are 00:17:19.359 --> 00:17:21.400 generated. Now, you'll definitely want to add 00:17:21.400 --> 00:17:24.000 your custom voice, right? Make it truly you. 00:17:24.160 --> 00:17:26.559 And that's where a tool like 11 Labs comes in. 00:17:26.559 --> 00:17:29.430 It helps you truly find your voice again in a 00:17:29.430 --> 00:17:32.089 digital sense. So this puts your actual voice 00:17:32.089 --> 00:17:35.190 or a custom one into the digital twin. Exactly. 00:17:35.849 --> 00:17:38.210 The process for voice cloning to get the best 00:17:38.210 --> 00:17:41.250 results is pretty specific. You'll want to find 00:17:41.250 --> 00:17:44.250 a quiet environment obviously. Use a good microphone. 00:17:44.529 --> 00:17:46.750 Even a decent smartphone mic in a quiet room 00:17:46.750 --> 00:17:49.710 can work okay to start. And then read about three 00:17:49.710 --> 00:17:51.990 to five minutes of text with a natural expressive 00:17:51.990 --> 00:17:54.930 tone. Don't be monotone. The quality of your 00:17:54.930 --> 00:17:57.450 input audio directly determines the quality of 00:17:57.450 --> 00:18:00.069 the cloned voice. Garbage in, garbage out applies 00:18:00.069 --> 00:18:02.170 here too. And what's really revolutionary you 00:18:02.170 --> 00:18:03.829 mentioned is the speech -to -speech feature. 00:18:04.250 --> 00:18:07.670 Yes. This is super cool. It analyzes an original 00:18:07.670 --> 00:18:09.869 audio file, maybe the synthesized one from your 00:18:09.869 --> 00:18:13.170 VEO3 video, to extract the melody of the speech. 00:18:13.430 --> 00:18:16.450 It's rhythm, the intonation, the pauses, all 00:18:16.450 --> 00:18:19.250 that natural cadence. It then applies that exact 00:18:19.250 --> 00:18:21.609 melody to the voice you've chosen, whether it's 00:18:21.609 --> 00:18:24.170 your own clone voice or maybe another synthesized 00:18:24.170 --> 00:18:27.130 one you like. The result is a new audio file 00:18:27.130 --> 00:18:29.529 with your voice, but speaking with the exact 00:18:29.529 --> 00:18:32.230 timing and cadence of the original video's audio 00:18:32.230 --> 00:18:34.539 track, it matches the lip sync perfectly. That 00:18:34.539 --> 00:18:36.640 is impressive. But it's important we touch on 00:18:36.640 --> 00:18:38.740 the crucial ethical issues here, right? Voice 00:18:38.740 --> 00:18:40.920 cloning technology is powerful, and it carries 00:18:40.920 --> 00:18:43.799 potential risks. Always, always only clone your 00:18:43.799 --> 00:18:45.460 own voice or the voice of someone who has given 00:18:45.460 --> 00:18:47.960 you clear, explicit permission. Using someone 00:18:47.960 --> 00:18:50.000 else's voice without their consent is unethical 00:18:50.000 --> 00:18:52.559 and, in many places, potentially illegal. We 00:18:52.559 --> 00:18:54.740 have to be responsible with this stuff. Absolutely 00:18:54.740 --> 00:18:56.640 crucial point, and once you've swapped in the 00:18:56.640 --> 00:18:59.099 voice using speech -to -speech, when you're dubbing 00:18:59.099 --> 00:19:01.339 and refining in your editing software, pay attention 00:19:01.339 --> 00:19:04.319 to ambient sounds. You might need to add small 00:19:04.319 --> 00:19:06.119 sound effects, the rustle of wind if they're 00:19:06.119 --> 00:19:08.339 outside, the click of a button they press, just 00:19:08.339 --> 00:19:10.779 little things to make the scene more lively and 00:19:10.779 --> 00:19:13.819 believable. Okay, final step. Assembling your 00:19:13.819 --> 00:19:16.339 complete scene. Let's talk about Google Flow. 00:19:17.019 --> 00:19:19.440 You should think of Google Flow as your AI rough 00:19:19.440 --> 00:19:22.519 cut desk. It's not a full -featured nonlinear 00:19:22.519 --> 00:19:25.319 editor like DaVinci Resolve or Premiere Pro. 00:19:25.500 --> 00:19:27.839 Definitely not. But it's the perfect place to 00:19:27.839 --> 00:19:30.180 quickly assemble your AI clips and just see if 00:19:30.180 --> 00:19:32.480 your story actually works, if the flow is right. 00:19:32.819 --> 00:19:35.539 So Flow is like a smart editor for your AI generated 00:19:35.539 --> 00:19:37.460 clips. Yeah, you could say that. It's a powerful 00:19:37.460 --> 00:19:40.259 preliminary assembly tool. Its strengths really 00:19:40.259 --> 00:19:43.039 lie in the ability to quickly preview and rearrange 00:19:43.039 --> 00:19:45.480 clips, and especially that feature to seamlessly 00:19:45.480 --> 00:19:48.099 extend a shot if you need it a bit longer. It's 00:19:48.099 --> 00:19:50.180 a powerful pre -production tool, really. Saves 00:19:50.180 --> 00:19:52.539 you significant time before you move on to more 00:19:52.539 --> 00:19:55.740 complex post -production. But it does have weaknesses. 00:19:56.200 --> 00:19:57.980 It definitely lacks advanced post -production 00:19:57.980 --> 00:20:00.980 tools like color grading, detailed audio mixing, 00:20:01.180 --> 00:20:04.460 or adding complex motion graphics. So the best 00:20:04.460 --> 00:20:07.470 workflow is probably Do your rough assembly and 00:20:07.470 --> 00:20:09.789 flow, check the timing and story, export the 00:20:09.789 --> 00:20:12.289 high quality clips, and then finalize everything 00:20:12.289 --> 00:20:14.750 in a professional non -linear editor like Resolve 00:20:14.750 --> 00:20:17.569 or Premiere. Stepping back, the profound impact 00:20:17.569 --> 00:20:20.049 of this whole revolution is, well, it's immense. 00:20:20.549 --> 00:20:22.650 It genuinely breaks down those huge financial 00:20:22.650 --> 00:20:24.869 barriers. It enables small businesses, independent 00:20:24.869 --> 00:20:27.029 creators, artists, even just curious individuals 00:20:27.029 --> 00:20:29.390 to produce ads and content with a quality that, 00:20:29.849 --> 00:20:32.329 honestly, can rival major brands sometimes. Think 00:20:32.329 --> 00:20:34.940 about it. You can A -B test creative ideas at 00:20:34.940 --> 00:20:37.339 breakneck speed, build a global personal brand 00:20:37.339 --> 00:20:39.380 from the comfort of your home office, all without 00:20:39.380 --> 00:20:41.759 the traditional overhead. It's a profound shift 00:20:41.759 --> 00:20:44.039 in how we create and share our messages. And 00:20:44.039 --> 00:20:46.559 this is truly just the beginning, isn't it? I 00:20:46.559 --> 00:20:49.460 mean, imagine a future where you can create interactive 00:20:49.460 --> 00:20:52.400 AI characters in real time, or produce cinema 00:20:52.400 --> 00:20:55.559 quality short films right on your laptop, or 00:20:55.559 --> 00:20:58.319 seamlessly integrate these digital actors into 00:20:58.319 --> 00:21:01.019 virtual production environments. We are really 00:21:01.019 --> 00:21:03.609 standing at the dawn of a whole new era of creativity, 00:21:04.029 --> 00:21:06.390 and it's just incredibly exciting to be watching 00:21:06.390 --> 00:21:09.690 it unfold, let alone participate in it. The technology 00:21:09.690 --> 00:21:12.809 is just moving so, so fast. The tools are here. 00:21:13.609 --> 00:21:15.289 The knowledge, hopefully, we've equipped you 00:21:15.289 --> 00:21:17.589 with some today. The next step is entirely yours. 00:21:17.769 --> 00:21:20.410 Don't be afraid to fail, really. Every experiment 00:21:20.410 --> 00:21:22.069 is a lesson learned, especially with this stuff. 00:21:22.349 --> 00:21:24.660 Start small. Try one thing, generate one image, 00:21:24.839 --> 00:21:27.299 make one short clip. See where your imagination 00:21:27.299 --> 00:21:29.940 combined with the power of AI can take you. Yeah. 00:21:30.019 --> 00:21:31.859 And the technology in this field is developing 00:21:31.859 --> 00:21:35.019 at just a dizzying pace. New features, new capabilities 00:21:35.019 --> 00:21:36.980 are being released constantly, it feels like. 00:21:37.359 --> 00:21:40.460 So consider following AI video communities, creators 00:21:40.460 --> 00:21:42.920 specializing in this area online. Just to stay 00:21:42.920 --> 00:21:44.619 updated on the latest developments, it changes 00:21:44.619 --> 00:21:47.279 almost weekly. So the provocative thought to 00:21:47.279 --> 00:21:49.700 leave you with is this. What will you create 00:21:49.700 --> 00:21:52.259 first with your digital twin? Thank you for joining 00:21:52.259 --> 00:21:54.819 us on this deep dive. O -U -T -R -O music.