WEBVTT 00:00:00.000 --> 00:00:02.100 Not long ago, a cinematic short film demanded 00:00:02.100 --> 00:00:04.960 a director. It needed a camera crew and a massive 00:00:04.960 --> 00:00:07.160 budget. You need a complex lighting knowledge 00:00:07.160 --> 00:00:10.199 for it to look acceptable. In 2026, it requires 00:00:10.199 --> 00:00:12.380 a standard laptop. You just need a solid prompt 00:00:12.380 --> 00:00:15.339 and some basic patience. And you need about 30 00:00:15.339 --> 00:00:18.789 minutes of focus time. Welcome to our deep dive 00:00:18.789 --> 00:00:20.809 today. We're unpacking a fascinating new workflow 00:00:20.809 --> 00:00:23.750 guide right now. It's called Mastering Text -to 00:00:23.750 --> 00:00:27.030 -Video AI in 2026. We're moving far past the 00:00:27.030 --> 00:00:29.170 early software hype today. We really want to 00:00:29.170 --> 00:00:30.890 unpack the actual daily production workflow. 00:00:30.989 --> 00:00:32.729 We're going to explore three core foundations 00:00:32.729 --> 00:00:35.750 of AI video. These foundations separate inconsistent 00:00:35.750 --> 00:00:38.829 outputs from true pro -level cinematic films. 00:00:39.390 --> 00:00:41.810 beat. It's just a massive shift in creative power. 00:00:41.969 --> 00:00:44.609 Yeah, it really is a completely new paradigm 00:00:44.609 --> 00:00:48.250 for creators. We're seeing an absolute explosion 00:00:48.250 --> 00:00:51.250 in AI video lately. Creators are producing stunning 00:00:51.250 --> 00:00:54.130 short films every single day. They're making 00:00:54.130 --> 00:00:56.210 high end branded ads right from their bedrooms. 00:00:56.270 --> 00:00:58.109 They do all of this without a single physical 00:00:58.109 --> 00:01:01.429 camera. The creative landscape has shifted dramatically 00:01:01.429 --> 00:01:04.609 in 12 short months. But there's a brutal truth 00:01:04.609 --> 00:01:08.049 we actually must face here. Most beginners fail 00:01:08.049 --> 00:01:11.170 completely when they try these new tools. It 00:01:11.170 --> 00:01:12.709 isn't because the video software is actually 00:01:12.709 --> 00:01:15.530 bad. It's because they jump in without a structured 00:01:15.530 --> 00:01:17.450 system. They don't understand how the underlying 00:01:17.450 --> 00:01:19.409 models actually work. I actually have to make 00:01:19.409 --> 00:01:21.730 a vulnerable admission right here. I still wrestle 00:01:21.730 --> 00:01:23.430 with burning through credits myself sometimes. 00:01:23.609 --> 00:01:25.609 I just roll the dice and hope for one decent 00:01:25.609 --> 00:01:27.689 clip. It can be an incredibly frustrating creative 00:01:27.689 --> 00:01:30.069 process, you know, but that makes a lot of sense 00:01:30.069 --> 00:01:32.150 when you step back. These tools are incredibly 00:01:32.150 --> 00:01:35.219 advanced today. So why is our instinct to just 00:01:35.219 --> 00:01:37.099 type a basic sentence? Why does hitting generate 00:01:37.099 --> 00:01:39.780 actually set us up to fail completely? Well, 00:01:39.840 --> 00:01:42.200 because surface level prompting lacks any real 00:01:42.200 --> 00:01:44.859 structural control. You're basically asking a 00:01:44.859 --> 00:01:47.400 statistical model to guess absolutely everything. 00:01:47.700 --> 00:01:49.780 It has to guess the complex lighting and the 00:01:49.780 --> 00:01:52.280 camera lens. It guesses the focal length and 00:01:52.280 --> 00:01:54.640 the local physics. The compute burden is just 00:01:54.640 --> 00:01:56.920 incredibly high for the model. It simply cannot 00:01:56.920 --> 00:01:58.959 hold all those intricate variables perfectly 00:01:58.959 --> 00:02:02.060 together. When you type text, you open an infinite 00:02:02.060 --> 00:02:05.459 possibility space. The model traverses a massive 00:02:05.459 --> 00:02:08.479 high dimensional latent space very quickly. It 00:02:08.479 --> 00:02:11.120 tries to resolve random digital noise into clear 00:02:11.120 --> 00:02:13.900 shapes. Doing that across time in a video is 00:02:13.900 --> 00:02:16.520 incredibly unstable. You need a multi -step workflow 00:02:16.520 --> 00:02:19.060 to guide the model safely. You have to actively 00:02:19.060 --> 00:02:22.120 constrain that infinite possibility space. So 00:02:22.120 --> 00:02:24.900 bad results come from missing workflows, not 00:02:24.900 --> 00:02:27.120 from broken tools. Right. And that brings us 00:02:27.120 --> 00:02:29.180 to the first core foundation today. We really 00:02:29.180 --> 00:02:31.319 have to move beyond just typing simple text prompts. 00:02:31.520 --> 00:02:33.759 The guide outlines five distinct methods in the 00:02:33.759 --> 00:02:35.900 new workflow. You have to understand these five 00:02:35.900 --> 00:02:38.520 core approaches very deeply. Let's walk through 00:02:38.520 --> 00:02:40.780 those five specific methods right now. Method 00:02:40.780 --> 00:02:43.180 one is the basic text to video we just discussed. 00:02:43.939 --> 00:02:47.360 This is still highly useful for simple establishing 00:02:47.360 --> 00:02:49.900 shots. You use it when you just need a quick 00:02:49.900 --> 00:02:52.680 static landscape. Method two is called image 00:02:52.680 --> 00:02:54.780 to video in the current guide. This is where 00:02:54.780 --> 00:02:57.199 you start with a static picture first. You lock 00:02:57.199 --> 00:02:59.319 in the exact composition before you ever animate 00:02:59.319 --> 00:03:02.080 it. Exactly. And that eliminates half the guesswork 00:03:02.080 --> 00:03:04.780 for the model immediately. Generating video from 00:03:04.780 --> 00:03:07.500 pure text is kind of like sculpting thick smoke. 00:03:07.780 --> 00:03:09.919 You might shape a perfect face for one single 00:03:09.919 --> 00:03:12.300 second, but the temporal wind immediately blows 00:03:12.300 --> 00:03:14.819 those delicate features away. An image prompt 00:03:14.819 --> 00:03:17.120 acts like a sudden flash freeze. It turns that 00:03:17.120 --> 00:03:19.879 highly unstable smoke into solid digital ice. 00:03:20.139 --> 00:03:22.680 You establish the lighting and the exact framing 00:03:22.680 --> 00:03:25.419 up front. Then you just ask the AI to add the 00:03:25.419 --> 00:03:28.639 subtle motion. Beat. That brings us to method 00:03:28.639 --> 00:03:30.840 three in the guide today. Method three is elements 00:03:30.840 --> 00:03:33.340 to video or ingredients to video, as some call 00:03:33.340 --> 00:03:35.979 it. Method four focuses specifically on generating 00:03:35.979 --> 00:03:38.860 accurate lip sync data. You absolutely need this 00:03:38.860 --> 00:03:41.280 if your characters are speaking dialogue. And 00:03:41.280 --> 00:03:43.599 method five is known as video to video or motion 00:03:43.599 --> 00:03:46.219 transfer. That means using real video movement 00:03:46.219 --> 00:03:49.680 to puppet an AI generated character. When I look 00:03:49.680 --> 00:03:51.879 at all these different methods combined, it feels 00:03:51.879 --> 00:03:54.020 like stacking Lego blocks of data to build a 00:03:54.020 --> 00:03:56.110 scene. It's so much better than hoping a single 00:03:56.110 --> 00:03:59.069 prompt works. That Lego analogy is the absolute 00:03:59.069 --> 00:04:01.150 perfect way to view it. You're basically assembling 00:04:01.150 --> 00:04:03.270 a final product from highly discrete pieces. 00:04:03.490 --> 00:04:05.430 Each individual piece serves a highly specific 00:04:05.430 --> 00:04:08.349 visual function. You don't ask one block to build 00:04:08.349 --> 00:04:10.969 the whole entire house. If you do, the AI model 00:04:10.969 --> 00:04:13.490 is going to hallucinate wildly. Yeah. So I want 00:04:13.490 --> 00:04:15.110 to ask about method number three specifically. 00:04:15.409 --> 00:04:17.529 How does moving to something like elements to 00:04:17.529 --> 00:04:19.730 video fundamentally change things? How does it 00:04:19.730 --> 00:04:21.910 change the control we have over the final shot? 00:04:22.319 --> 00:04:24.620 Well, separating visual elements allows you to 00:04:24.620 --> 00:04:27.519 provide precise direction. You isolate the visual 00:04:27.519 --> 00:04:30.399 variables completely away from each other. If 00:04:30.399 --> 00:04:32.420 you generate a character and a background together, 00:04:32.560 --> 00:04:35.899 they bleed. The AI struggles to understand where 00:04:35.899 --> 00:04:38.540 the person actually ends. Right. It lacks true 00:04:38.540 --> 00:04:42.079 spatial awareness of the digital scene. It might 00:04:42.079 --> 00:04:44.240 just blend a green shirt into a green forest. 00:04:44.540 --> 00:04:47.319 By splitting them apart, you isolate the visual 00:04:47.319 --> 00:04:49.959 variables completely. You generate your human 00:04:49.959 --> 00:04:52.500 character on a solid green screen. You generate 00:04:52.500 --> 00:04:54.680 your highly detailed forest background entirely 00:04:54.680 --> 00:04:57.100 separately. Then you composite them together 00:04:57.100 --> 00:04:59.399 in a traditional video editor. Mixing separate 00:04:59.399 --> 00:05:01.740 ingredients gives us total control over the final 00:05:01.740 --> 00:05:04.680 visual recipe. That's exactly how the top professionals 00:05:04.680 --> 00:05:07.120 are working right now. But having those five 00:05:07.120 --> 00:05:09.339 methods is pretty much useless without visual 00:05:09.339 --> 00:05:11.699 consistency. It doesn't matter if you can move 00:05:11.699 --> 00:05:14.339 a character smoothly. If their face changes every 00:05:14.339 --> 00:05:17.019 time the camera cuts, you fail. The illusion 00:05:17.019 --> 00:05:19.439 of your cinematic short film is instantly broken. 00:05:19.800 --> 00:05:22.600 We call this frustrating phenomenon the cousin 00:05:22.600 --> 00:05:25.759 effect. Your hero suddenly looks like their own 00:05:25.759 --> 00:05:28.819 slightly weird cousin. This brings us directly 00:05:28.819 --> 00:05:31.660 to Foundation 2, which is the consistency playbook. 00:05:32.089 --> 00:05:33.850 Consistency is definitely where beginners usually 00:05:33.850 --> 00:05:36.449 lose their minds entirely. You get one truly 00:05:36.449 --> 00:05:39.050 great shot of your main hero, then the next shot 00:05:39.050 --> 00:05:40.990 looks like a completely different person. Yeah, 00:05:41.069 --> 00:05:43.870 it is incredibly jarring for the viewer. The 00:05:43.870 --> 00:05:46.610 guide gives us specific workflow steps to prevent 00:05:46.610 --> 00:05:49.389 these mistakes. First, you must always build 00:05:49.389 --> 00:05:51.589 from static images first. You shouldn't ever 00:05:51.589 --> 00:05:54.629 start your complex workflow with a video. I guess 00:05:54.629 --> 00:05:56.610 because images are just incredibly cheap. They're 00:05:56.610 --> 00:05:58.649 very fast to generate compared to full video. 00:05:59.079 --> 00:06:01.740 Exactly. You can iterate on a face dozens of 00:06:01.740 --> 00:06:04.579 times really quickly. You can explore a vast 00:06:04.579 --> 00:06:07.680 latent space without burning your budget. Video 00:06:07.680 --> 00:06:10.019 generation takes longer and costs significantly 00:06:10.019 --> 00:06:12.420 more compute credits. You really want to lock 00:06:12.420 --> 00:06:14.240 in the still image before spending resources. 00:06:14.800 --> 00:06:17.980 Second, you use your AI generated output as the 00:06:17.980 --> 00:06:20.279 new reference point. You feed the good results 00:06:20.279 --> 00:06:22.779 back into the system continually. You essentially 00:06:22.779 --> 00:06:26.120 lock the visual seed in place permanently. Third, 00:06:26.560 --> 00:06:28.819 You must generate environments completely separately 00:06:28.819 --> 00:06:31.240 from the characters. We touched on this briefly 00:06:31.240 --> 00:06:34.180 during the isolated ingredients discussion. Fourth, 00:06:34.379 --> 00:06:36.759 you have to build a master character reference 00:06:36.759 --> 00:06:39.540 sheet. this is a very specific type of foundational 00:06:39.540 --> 00:06:42.300 document right and this is the absolute secret 00:06:42.300 --> 00:06:45.360 weapon of pro creators you generate a grid showing 00:06:45.360 --> 00:06:47.339 your character from multiple distinct angles 00:06:47.339 --> 00:06:49.439 you want a clear front profile a side profile 00:06:49.439 --> 00:06:52.439 and a back you upscale this image grid to lock 00:06:52.439 --> 00:06:54.360 in the micro details then you feed this grid 00:06:54.360 --> 00:06:56.720 back into your image comp continually you set 00:06:56.720 --> 00:06:58.699 the character weight extremely high in the software 00:06:59.149 --> 00:07:02.009 This forces the generative model to anchor onto 00:07:02.009 --> 00:07:04.709 those specific pixels. And finally, you apply 00:07:04.709 --> 00:07:08.050 this exact same logic to scene props. It also 00:07:08.050 --> 00:07:09.730 applies to any supporting characters in your 00:07:09.730 --> 00:07:15.089 short film. Beat. Whoa. Beat. Imagine generating 00:07:15.089 --> 00:07:18.230 an entire persistent cinematic universe from 00:07:18.230 --> 00:07:20.839 just one character sheet. It really is a staggering 00:07:20.839 --> 00:07:23.240 level of creative power today. You're essentially 00:07:23.240 --> 00:07:25.720 building a digital backlot on your personal computer. 00:07:25.939 --> 00:07:28.339 You lock in the exact visual identity before 00:07:28.339 --> 00:07:31.240 you ever animate anything. This stops the generative 00:07:31.240 --> 00:07:33.439 model from hallucinating completely new details. 00:07:33.660 --> 00:07:35.740 It has a concrete visual anchor to reference 00:07:35.740 --> 00:07:38.040 constantly during generation. You completely 00:07:38.040 --> 00:07:40.040 eliminate that terrifying cousin effect we mentioned 00:07:40.040 --> 00:07:42.500 earlier. I need to dig into one specific workflow 00:07:42.500 --> 00:07:45.680 step right here. Why is it so critical to generate 00:07:45.680 --> 00:07:48.160 environments completely separately from our characters? 00:07:48.730 --> 00:07:51.069 Because AI models really struggle with complex 00:07:51.069 --> 00:07:53.829 spatial relationships natively, they don't truly 00:07:53.829 --> 00:07:56.310 understand three -dimensional depth like humans 00:07:56.310 --> 00:07:59.199 actually do. If you ask for a man in a bustling 00:07:59.199 --> 00:08:01.800 neon city, the model might start putting neon 00:08:01.800 --> 00:08:04.360 lights directly on his leather jacket. It gets 00:08:04.360 --> 00:08:06.740 completely confused by all the overlapping digital 00:08:06.740 --> 00:08:09.459 noise. Isolating backgrounds stops the AI from 00:08:09.459 --> 00:08:11.439 blending everything into total visual chaos. 00:08:11.779 --> 00:08:15.100 Exactly. You handle the complex compositing logic 00:08:15.100 --> 00:08:18.500 yourself later on. You don't force the AI to 00:08:18.500 --> 00:08:21.660 do complex spatial math. Okay, let's step back 00:08:21.660 --> 00:08:23.689 and look at where we are now. We have highly 00:08:23.689 --> 00:08:25.990 consistent characters and fully locked in visual 00:08:25.990 --> 00:08:28.550 environments. We have our master character sheets 00:08:28.550 --> 00:08:31.410 completely ready to go. The latent space is tightly 00:08:31.410 --> 00:08:34.090 constrained and fully controlled. So how do we 00:08:34.090 --> 00:08:36.090 actually get them to move naturally? Foundation 00:08:36.090 --> 00:08:38.470 3 says we have to prompt like a film director. 00:08:38.730 --> 00:08:40.789 We cannot just prompt like a software engineer 00:08:40.789 --> 00:08:42.950 anymore. We have to think about blocking, pacing, 00:08:43.129 --> 00:08:46.269 and direct action. This is where the true cinematic 00:08:46.269 --> 00:08:48.870 magic really starts happening. The guide breaks 00:08:48.870 --> 00:08:51.870 down three specific directing rules for us. The 00:08:51.870 --> 00:08:53.730 first rule is all about controlling the pace 00:08:53.730 --> 00:08:56.740 of motion. You must use the word slow in your 00:08:56.740 --> 00:08:59.100 text prompts. You have to use it way more than 00:08:59.100 --> 00:09:01.159 you think you need to. That feels incredibly 00:09:01.159 --> 00:09:03.480 counterintuitive at first glance. You know, you 00:09:03.480 --> 00:09:05.940 want dynamic action, so you naturally ask for 00:09:05.940 --> 00:09:08.620 fast movement. But the physics engines in AI 00:09:08.620 --> 00:09:11.399 naturally want to move things way too fast. Right. 00:09:11.399 --> 00:09:13.600 And they struggle deeply with subtle temporal 00:09:13.600 --> 00:09:17.259 consistency. If you ask a character to run quickly 00:09:17.259 --> 00:09:20.340 across a room, the model has to invent a massive 00:09:20.340 --> 00:09:23.340 amount of new pixel data rapidly. It usually 00:09:23.340 --> 00:09:26.360 fails and creates a blurry, chaotic visual mess. 00:09:26.960 --> 00:09:29.840 Asking for slow motion forces the engine to calculate 00:09:29.840 --> 00:09:32.740 carefully. It gives the AI time to maintain the 00:09:32.740 --> 00:09:34.860 rigid structural integrity. You can always speed 00:09:34.860 --> 00:09:37.120 the footage up in post -production later. Yeah. 00:09:37.360 --> 00:09:39.259 Beat, that makes perfect sense when you understand 00:09:39.259 --> 00:09:41.659 mechanics. So what is the second critical directing 00:09:41.659 --> 00:09:44.210 rule from the guide? You must restate what is 00:09:44.210 --> 00:09:47.129 already visible in the reference image. If your 00:09:47.129 --> 00:09:49.429 character is wearing a heavy red leather jacket, 00:09:49.690 --> 00:09:52.129 you have to type red leather jacket in the animation 00:09:52.129 --> 00:09:55.289 prompt. You cannot just say make the man walk 00:09:55.289 --> 00:09:58.110 forward. You have to consciously remind the AI 00:09:58.110 --> 00:10:01.009 about the specific clothing. If I feed the AI 00:10:01.009 --> 00:10:04.269 an image, why waste prompt space restating visible 00:10:04.269 --> 00:10:07.120 details? Well, because the generative model has 00:10:07.120 --> 00:10:09.700 a strictly finite attention budget. Think of 00:10:09.700 --> 00:10:12.000 the AI like a highly stressed camera operator. 00:10:12.299 --> 00:10:15.299 It only has so much compute power for every single 00:10:15.299 --> 00:10:18.779 frame. Calculating complex motion vectors is 00:10:18.779 --> 00:10:21.179 mathematically very expensive for the system. 00:10:21.360 --> 00:10:23.840 It has to calculate exactly how an arm swings 00:10:23.840 --> 00:10:26.960 through space. Exactly. As it calculates the 00:10:26.960 --> 00:10:30.179 complex math of a moving arm. It spends 90 % 00:10:30.179 --> 00:10:32.240 of its attention budget right there. It sometimes 00:10:32.240 --> 00:10:34.200 drops the data about the red jacket entirely. 00:10:34.679 --> 00:10:37.419 It prioritizes the new motion vector over the 00:10:37.419 --> 00:10:40.440 existing visual stability. It'll literally hallucinate 00:10:40.440 --> 00:10:43.379 a generic gray t -shirt to save processing power. 00:10:43.659 --> 00:10:45.220 It just completely forgets what the character 00:10:45.220 --> 00:10:47.470 is currently wearing. Yeah, it takes the path 00:10:47.470 --> 00:10:50.909 of least computational resistance. Restating 00:10:50.909 --> 00:10:54.149 the details forces the AI to actively focus during 00:10:54.149 --> 00:10:57.409 generation. It preserves those specific visual 00:10:57.409 --> 00:11:00.230 elements during the actual motion. You essentially 00:11:00.230 --> 00:11:02.929 force it to allocate its attention budget properly. 00:11:03.470 --> 00:11:06.570 Repeating visual details forces the AI to remember 00:11:06.570 --> 00:11:09.509 exactly what it's animating. Precisely. You become 00:11:09.509 --> 00:11:12.190 a strict director managing a stressed camera 00:11:12.190 --> 00:11:14.600 crew. And that brings us to the third critical 00:11:14.600 --> 00:11:17.279 directing rule today. You must stick to very 00:11:17.279 --> 00:11:20.559 simple, highly isolated physical actions. You 00:11:20.559 --> 00:11:23.000 only want one or two clear actions per shot. 00:11:23.159 --> 00:11:25.379 You don't ask the character to walk, talk, and 00:11:25.379 --> 00:11:27.639 drink coffee simultaneously. The attention budget 00:11:27.639 --> 00:11:30.320 would just completely crash. Exactly. The AI 00:11:30.320 --> 00:11:32.480 model would panic and melt the image entirely. 00:11:32.759 --> 00:11:34.700 You have to break those complex actions down 00:11:34.700 --> 00:11:37.240 into individual camera shots. Shot one is just 00:11:37.240 --> 00:11:38.899 the character walking slowly through the door. 00:11:39.080 --> 00:11:41.240 Shot two is a tight close -up of them taking 00:11:41.240 --> 00:11:44.049 a sip of coffee. Shot 3 is them delivering the 00:11:44.049 --> 00:11:46.610 actual spoken dialogue. This drastically reduces 00:11:46.610 --> 00:11:49.029 the cognitive load on the AI model. It ensures 00:11:49.029 --> 00:11:51.370 each individual segment looks as realistic as 00:11:51.370 --> 00:11:53.720 mathematically possible. Right, and that's exactly 00:11:53.720 --> 00:11:56.340 how real film directors work anyway. They build 00:11:56.340 --> 00:11:58.700 complex scenes through a sequence of highly controlled 00:11:58.700 --> 00:12:01.159 shots. They don't capture the entire movie in 00:12:01.159 --> 00:12:03.679 one chaotic take. They manage the variables on 00:12:03.679 --> 00:12:06.379 set to guarantee a perfect outcome. By limiting 00:12:06.379 --> 00:12:08.480 the variables in your prompt, you guarantee a 00:12:08.480 --> 00:12:10.820 much higher success rate. You basically stop 00:12:10.820 --> 00:12:12.820 rolling the dice and hoping for a lucky generation. 00:12:13.399 --> 00:12:16.019 Which brings us to the massive overarching takeaway 00:12:16.019 --> 00:12:20.500 from all this. The 2026 AI Video Gold Rush. isn't 00:12:20.500 --> 00:12:23.340 about the software you use. Every single creator 00:12:23.340 --> 00:12:25.840 essentially has access to the exact same foundational 00:12:25.840 --> 00:12:29.080 tools. The latent space is available to absolutely 00:12:29.080 --> 00:12:31.559 everyone with a laptop. It's really about the 00:12:31.559 --> 00:12:33.759 intense discipline of your daily creative workflow. 00:12:34.039 --> 00:13:00.519 You stop relying on lucky generations and magical 00:13:00.519 --> 00:13:04.120 one -shot prompts. Beat. You essentially transform 00:13:04.120 --> 00:13:06.639 from a prompt guesser into an actual director. 00:13:06.899 --> 00:13:09.320 You take total control of the underlying mathematical 00:13:09.320 --> 00:13:11.659 system. And that transformation is what truly 00:13:11.659 --> 00:13:14.500 separates the amateurs from the pros. Anyone 00:13:14.500 --> 00:13:17.159 can generate a random, morphing, five -second 00:13:17.159 --> 00:13:20.980 video clip today, but directing a cohesive, visually 00:13:20.980 --> 00:13:23.639 perfect narrative requires a deeply structured 00:13:23.639 --> 00:13:26.820 system. It requires immense patience, careful 00:13:26.820 --> 00:13:29.100 planning, and a deep understanding of the medium. 00:13:29.320 --> 00:13:31.919 You have to understand latent space, attention 00:13:31.919 --> 00:13:35.139 budgets, and complex motion vectors. It's a completely 00:13:35.139 --> 00:13:38.220 new craft that demands serious artistic respect. 00:13:38.759 --> 00:13:40.500 Thank you so much for joining us on this deep 00:13:40.500 --> 00:13:42.360 dive. We want to leave you with a very simple 00:13:42.360 --> 00:13:44.379 challenge today. Try taking just one of these 00:13:44.379 --> 00:13:46.700 new structured workflow steps. Try building a 00:13:46.700 --> 00:13:48.820 solid multi -angle character reference sheet 00:13:48.820 --> 00:13:51.340 first. Do this before you ever hit generate video 00:13:51.340 --> 00:13:54.179 again. Build that crucial visual foundation before 00:13:54.179 --> 00:13:56.480 you try to build a whole house. Lock down your 00:13:56.480 --> 00:13:58.419 digital actor before you force them to perform. 00:13:58.679 --> 00:14:01.139 It'll truly save you so much frustration and 00:14:01.139 --> 00:14:03.159 wasted time. You'll finally feel like you're 00:14:03.159 --> 00:14:04.919 actually in control of the creative process. 00:14:05.200 --> 00:14:07.539 You'll stop fighting the AI and start directing 00:14:07.539 --> 00:14:10.460 it properly. If anyone can manifest a perfect 00:14:10.460 --> 00:14:13.919 cinematic short in 30 minutes, two sec silence. 00:14:14.259 --> 00:14:17.019 It leaves you wondering, when technical barriers 00:14:17.019 --> 00:14:19.460 completely vanish, will human taste be the only 00:14:19.460 --> 00:14:21.460 thing left? Will that be what makes a director 00:14:21.460 --> 00:14:23.240 valuable? Something to think about.