WEBVTT 00:00:00.000 --> 00:00:02.839 You can spend thousands of dollars on AI video 00:00:02.839 --> 00:00:04.799 generation right now, and what you get back is, 00:00:04.900 --> 00:00:08.560 well, what experts are calling AI slop? Oh, yeah. 00:00:09.519 --> 00:00:12.599 Unusable, blurry junk. It looks worse than, I 00:00:12.599 --> 00:00:15.779 don't know, early 2000s computer graphics. And 00:00:15.779 --> 00:00:18.140 that one core decision, which brain you choose 00:00:18.140 --> 00:00:20.820 for your shot, that's the absolute difference 00:00:20.820 --> 00:00:23.480 between success and just utter frustration. It 00:00:23.480 --> 00:00:25.460 changes everything. It changes your budget, your 00:00:25.460 --> 00:00:28.219 outcome. You really need to stop describing things 00:00:28.219 --> 00:00:30.129 and start directing them. Welcome to the Deep 00:00:30.129 --> 00:00:32.770 Dive. Today, we are taking a look at the AI video 00:00:32.770 --> 00:00:36.289 landscape as it stands in early 2026. We've got 00:00:36.289 --> 00:00:38.810 an expert ranking of the top 10 models. And this 00:00:38.810 --> 00:00:41.070 is really the ultimate guide to saving you time, 00:00:41.130 --> 00:00:43.770 money, and I think most importantly, your sanity 00:00:43.770 --> 00:00:45.950 in what has become a completely overwhelming 00:00:45.950 --> 00:00:48.869 field. Yeah, our mission here is pretty surgical. 00:00:48.990 --> 00:00:50.689 We want to cut right through the marketing hype. 00:00:50.990 --> 00:00:53.049 We're going to establish which tools are actually 00:00:53.049 --> 00:00:56.729 dominating, like Cling 2 .6, Sora 2, and which 00:00:56.729 --> 00:00:58.590 of the older leaders, you know, like Runway. 00:00:58.939 --> 00:01:01.140 They've become seriously over -Christ disappointments. 00:01:01.159 --> 00:01:04.420 You're paying for brand loyalty that just isn't 00:01:04.420 --> 00:01:07.159 earned anymore. And central to this whole conversation 00:01:07.159 --> 00:01:10.500 is this new paradigm that seems to separate the 00:01:10.500 --> 00:01:13.340 pros from the hobbyists. It's all about orchestration. 00:01:13.780 --> 00:01:16.099 Orchestration. So define that for us. Orchestration 00:01:16.099 --> 00:01:18.519 is everything now. It means you're not just, 00:01:18.560 --> 00:01:20.739 you know, typing in a description and hoping 00:01:20.739 --> 00:01:23.560 for the best. Right. You are actively directing 00:01:23.560 --> 00:01:26.799 the AI, its motion, its character, its composition. 00:01:27.739 --> 00:01:30.780 All using these advanced inputs. Like what kind 00:01:30.780 --> 00:01:32.780 of inputs? Things like locked reference images 00:01:32.780 --> 00:01:37.480 and really specific start and end frames to guide 00:01:37.480 --> 00:01:39.799 the movement. Okay, let's unpack this then. I 00:01:39.799 --> 00:01:42.239 mean, the field is just busying. We've got, what, 00:01:42.319 --> 00:01:44.780 10 plus platforms all shouting that they're the 00:01:44.780 --> 00:01:47.040 new leader? They all say it. How do you even 00:01:47.040 --> 00:01:50.060 begin to choose the right brain for a project? 00:01:50.579 --> 00:01:53.640 You have to start by defining quality. But in 00:01:53.640 --> 00:01:56.760 a really rigorous way, based purely on the output. 00:01:57.209 --> 00:01:58.849 Not the marketing. Not the marketing. So we're 00:01:58.849 --> 00:02:01.469 using a clear tier system based on real -world 00:02:01.469 --> 00:02:03.890 results. A tier is that mind -blowing quality. 00:02:03.950 --> 00:02:06.370 It's non -negotiable for serious work. Right. 00:02:06.489 --> 00:02:10.710 B tier is reliable, solid. C tier, that's where 00:02:10.710 --> 00:02:13.250 the tools get risky or fundamentally flawed. 00:02:13.389 --> 00:02:16.009 And D tier. You skip it entirely. That clear 00:02:16.009 --> 00:02:18.710 system right there, it instantly protects you 00:02:18.710 --> 00:02:21.750 from paying for a glossy logo. So to expose those... 00:02:22.699 --> 00:02:24.840 uncomfortable truths the sources we analyzed 00:02:24.840 --> 00:02:27.900 used a standardized test prompt across all the 00:02:27.900 --> 00:02:30.280 models yeah and it's a great visual example it's 00:02:30.280 --> 00:02:33.699 a scientist in a low -light research lab recoils 00:02:33.699 --> 00:02:36.360 in disbelief stumbling backward and crashing 00:02:36.360 --> 00:02:38.599 to the floor as an otherworldly creature slowly 00:02:38.599 --> 00:02:41.400 escapes a glass containment vessel right and 00:02:41.400 --> 00:02:43.479 what that test prompt immediately reveals is 00:02:43.479 --> 00:02:46.039 that brand recognition means absolutely nothing 00:02:46.039 --> 00:02:48.400 for fidelity it's not a guarantee not at all 00:02:48.400 --> 00:02:50.599 the most expensive tools often fail the moment 00:02:50.599 --> 00:02:52.490 the character needs to do something complex, 00:02:52.530 --> 00:02:55.229 awaited motion like stumbling and falling. We 00:02:55.229 --> 00:02:57.689 need substance, not just a flashy website. So 00:02:57.689 --> 00:03:00.270 what's the fundamental quality that separates 00:03:00.270 --> 00:03:02.870 A -tier from everything else in this rapidly 00:03:02.870 --> 00:03:05.629 changing market? It's photorealistic detail and 00:03:05.629 --> 00:03:07.930 consistency. The video has to look like real 00:03:07.930 --> 00:03:10.030 high -end footage, not something from a game 00:03:10.030 --> 00:03:12.210 engine. That brings us to the champions then, 00:03:12.330 --> 00:03:15.870 the A -tier tools, the ones you use when quality 00:03:15.870 --> 00:03:19.289 is absolutely non -negotiable. Kling AI 2 .6 00:03:19.289 --> 00:03:21.990 is the reigning champion right now. If you were 00:03:21.990 --> 00:03:25.770 forced to pick just one tool for a professional 00:03:25.770 --> 00:03:29.229 workflow, this is it. Why? What makes it win? 00:03:29.449 --> 00:03:33.229 In a word, clarity, visual clarity. It avoids 00:03:33.229 --> 00:03:36.250 that dreaded plastic filter effect. Okay, what 00:03:36.250 --> 00:03:38.469 exactly is that plastic filter effect we see 00:03:38.469 --> 00:03:40.349 everywhere else? It's what happens when the AI 00:03:40.349 --> 00:03:43.830 tries to smooth out artifacts or fix the jiggling. 00:03:43.949 --> 00:03:46.370 Temporal coherence issues. Exactly. It ends up 00:03:46.370 --> 00:03:49.129 softening everything. Clang avoids that. It preserves 00:03:49.129 --> 00:03:52.259 the tiny details. Skin textures, dust particles 00:03:52.259 --> 00:03:55.080 in the air, complex reflections on glass. The 00:03:55.080 --> 00:03:57.639 fine weave on a jacket. All of it. And the characters 00:03:57.639 --> 00:04:00.039 move like actual humans, not like interpolated 00:04:00.039 --> 00:04:01.939 video game characters. But the source has noted 00:04:01.939 --> 00:04:04.500 something in Kling that sounds truly revolutionary. 00:04:04.800 --> 00:04:06.979 It deals with sound right out of the box. This 00:04:06.979 --> 00:04:09.340 is the killer feature. Native audio. Native audio. 00:04:09.460 --> 00:04:11.639 It means it generates high -fidelity sound effects, 00:04:11.759 --> 00:04:14.340 even dialogue, synchronized right inside the 00:04:14.340 --> 00:04:17.079 video output. Before, you'd have to hire a sound 00:04:17.079 --> 00:04:19.800 designer or spend hours in an editor. Kling does 00:04:19.800 --> 00:04:22.819 the heavy lifting. So what's the catch? The only 00:04:22.819 --> 00:04:25.540 real constraint is that physics can occasionally 00:04:25.540 --> 00:04:27.819 break. You might see a character slide through 00:04:27.819 --> 00:04:30.339 a wall in maybe one out of eight generations. 00:04:30.720 --> 00:04:32.819 Manageable, I suppose, if you're saving days 00:04:32.819 --> 00:04:35.180 of post -production. Totally. And then we have 00:04:35.180 --> 00:04:39.860 Sora 2, the social media king. This tool is stunning, 00:04:39.920 --> 00:04:41.779 but the sources mentioned a real struggle point. 00:04:41.860 --> 00:04:45.339 They called it the walls. Yes, Sora is. It's 00:04:45.339 --> 00:04:48.139 polarizing because when it works, the quality 00:04:48.139 --> 00:04:52.560 is truly stunning. It's perfect for short form 00:04:52.560 --> 00:04:55.620 influencer style clips, stuff that looks native 00:04:55.620 --> 00:05:00.139 to TikTok or reels. feels like raw, high -end 00:05:00.139 --> 00:05:02.079 phone footage. But the walls... The walls are 00:05:02.079 --> 00:05:05.220 the aggressive content filters, the prompt rejections. 00:05:05.220 --> 00:05:07.420 They just feel arbitrary, and they completely 00:05:07.420 --> 00:05:09.939 break your creative flow. That sounds incredibly 00:05:09.939 --> 00:05:12.439 exhausting. It is. And what's worse, the image 00:05:12.439 --> 00:05:14.800 -to -video limitations are crippling for pro 00:05:14.800 --> 00:05:17.800 users. You can't upload images with people in 00:05:17.800 --> 00:05:20.279 them. It's strictly off -limits, which eliminates 00:05:20.279 --> 00:05:23.040 all the crucial consistency hacks we rely on 00:05:23.040 --> 00:05:26.170 for character work. So the Sora's stunning quality 00:05:26.170 --> 00:05:28.790 ultimately outweigh the exhaustion from all those 00:05:28.790 --> 00:05:31.910 content restrictions. Yes, but only for that 00:05:31.910 --> 00:05:35.569 niche, short -form viral content. The restrictions 00:05:35.569 --> 00:05:38.170 really limit its professional applications. Okay, 00:05:38.230 --> 00:05:40.649 let's move down to the B tier. These are the 00:05:40.649 --> 00:05:44.490 solid, reliable performers. Tools that are predictable 00:05:44.490 --> 00:05:47.889 and often just a much better value. Starting 00:05:47.889 --> 00:05:51.769 with Google VEO 3 .1. The sources call this the 00:05:51.769 --> 00:05:53.629 versatile workhorse. Workhorse, I like that. 00:05:53.949 --> 00:05:57.610 VEO is perfect for dynamic projects. One day 00:05:57.610 --> 00:06:00.110 it's sci -fi, the next it's a talking head product 00:06:00.110 --> 00:06:02.550 demo. So if your project has characters talking, 00:06:02.750 --> 00:06:05.449 VEO is the first choice. Absolutely. It has the 00:06:05.449 --> 00:06:07.589 most realistic facial expressions and lip syncing 00:06:07.589 --> 00:06:09.629 in the industry. Dialogue scenes actually feel 00:06:09.629 --> 00:06:12.370 safe here, which is... which is rare and critically 00:06:12.370 --> 00:06:15.689 veo offers robust camera control which ties right 00:06:15.689 --> 00:06:17.769 back to our idea of orchestration yeah what's 00:06:17.769 --> 00:06:20.250 fascinating here is that veo lets you force precise 00:06:20.250 --> 00:06:21.889 camera movement so you're not just describing 00:06:21.889 --> 00:06:25.189 a zoom no you upload a start frame and an end 00:06:25.189 --> 00:06:27.430 frame you're literally telling it start here 00:06:27.430 --> 00:06:32.350 end here rotations zooms pans they become predictable 00:06:32.350 --> 00:06:35.930 instead of a guess where does it fall short the 00:06:35.930 --> 00:06:40.139 main caveat is complex motion so falls or fast 00:06:40.139 --> 00:06:42.959 physical action can still look a little rubbery. 00:06:42.980 --> 00:06:45.019 And there's a quality and cost split too, right? 00:06:45.319 --> 00:06:48.079 Absolutely. The high quality mode is great, costs 00:06:48.079 --> 00:06:50.800 around $1 .25. But if you switch to standard 00:06:50.800 --> 00:06:54.360 mode, the output is notably softer. For how much? 00:06:54.500 --> 00:06:57.240 Just 25 cents. So you have to decide if that 00:06:57.240 --> 00:06:59.060 quarter is worth the clarity. And then we have 00:06:59.060 --> 00:07:02.120 the sleeper hit, Sedans 1 .5 Pro, which the sources 00:07:02.120 --> 00:07:05.019 have crowned the budget king. Sedans is the tool 00:07:05.019 --> 00:07:07.519 for when volume matters. It offers a sharpness 00:07:07.519 --> 00:07:10.329 that genuinely rivals Kling. Wow. At half the 00:07:10.329 --> 00:07:13.189 cost. It's around 52 cents per video. It delivers 00:07:13.189 --> 00:07:15.949 really strong visuals at a price that makes scaling 00:07:15.949 --> 00:07:18.490 for big campaigns realistic. They even bundle 00:07:18.490 --> 00:07:20.949 in basic audio. But what's the trade -off? For 00:07:20.949 --> 00:07:22.829 being the budget king, where do you sacrifice 00:07:22.829 --> 00:07:25.990 quality? You sacrifice control. Seedins kind 00:07:25.990 --> 00:07:28.149 of likes to move itself. What do you mean? Even 00:07:28.149 --> 00:07:30.350 with static comps, you often get camera drift. 00:07:30.529 --> 00:07:33.050 And the results can vary a lot between generations. 00:07:33.370 --> 00:07:36.050 You're trading that consistency in precise direction 00:07:36.050 --> 00:07:40.220 for volume and a low price. So if VEO is the 00:07:40.220 --> 00:07:42.860 best for talking heads, what's the biggest risk 00:07:42.860 --> 00:07:45.040 when you're choosing the cheaper sedans? You 00:07:45.040 --> 00:07:47.939 sacrifice motion control and consistency. You're 00:07:47.939 --> 00:07:49.959 going to experience camera drift and pretty varied 00:07:49.959 --> 00:07:51.800 outputs. All right, let's get into the C tier. 00:07:52.000 --> 00:07:54.519 These are the tools that require real caution. 00:07:54.939 --> 00:07:58.839 And we have to talk about RunwayGen 4 .5. Ah, 00:07:59.120 --> 00:08:02.420 Runway. A prime example of paying for the past 00:08:02.420 --> 00:08:05.000 instead of the present. It's so true. Runway 00:08:05.000 --> 00:08:06.939 was the industry leader just six months ago. 00:08:07.449 --> 00:08:10.170 It was known for smooth, intentional camera tracking. 00:08:10.610 --> 00:08:13.529 But now, it's severely overpriced. How much? 00:08:13.810 --> 00:08:16.810 About $2 .50 a clip. Yeah. and it completely 00:08:16.810 --> 00:08:19.730 underperforms its peers in quality. It seems 00:08:19.730 --> 00:08:21.949 like they've just been dethroned by a failure 00:08:21.949 --> 00:08:24.250 to maintain character consistency. Exactly. When 00:08:24.250 --> 00:08:26.550 it breaks, it breaks hard. You'll get amazing 00:08:26.550 --> 00:08:29.970 camera work, but up close, bodies bend unnaturally, 00:08:29.990 --> 00:08:33.649 hands warp, and faces just drift into that uncanny 00:08:33.649 --> 00:08:37.029 territory. You know, honestly, this is the kind 00:08:37.029 --> 00:08:39.169 of inconsistency I still wrestle with myself. 00:08:39.659 --> 00:08:41.879 You remember that one great tracking shot, but 00:08:41.879 --> 00:08:43.840 then the characters start looking strange up 00:08:43.840 --> 00:08:46.279 close and you have to discard the clip anyway. 00:08:46.519 --> 00:08:48.840 It's so frustrating. And the work slow itself 00:08:48.840 --> 00:08:50.820 is outdated, which is maybe why they fell so 00:08:50.820 --> 00:08:53.559 fast. That's a key piece of it. It's still mostly 00:08:53.559 --> 00:08:57.039 text to video only. No image to video. No. Which 00:08:57.039 --> 00:08:59.620 eliminates that necessary workflow the A and 00:08:59.620 --> 00:09:02.580 B tiers rely on for consistency. They just failed 00:09:02.580 --> 00:09:05.240 to integrate that modern method. No native audio 00:09:05.240 --> 00:09:07.740 either, so more work in post. It's a tool that 00:09:07.740 --> 00:09:10.340 feels stuck in the past. And you are definitely 00:09:10.340 --> 00:09:13.100 paying a premium for that brand loyalty. Okay, 00:09:13.159 --> 00:09:15.600 let's briefly cover the other C tier tools because 00:09:15.600 --> 00:09:17.960 they illustrate some pretty specific compromises. 00:09:18.320 --> 00:09:21.299 Lumire 3 is an interesting one. It excels at 00:09:21.299 --> 00:09:24.340 static landscapes. And it follows complex prompts 00:09:24.340 --> 00:09:26.919 really closely. But the second you introduce 00:09:26.919 --> 00:09:29.899 subtle motion, faces and mouths start to twitch. 00:09:30.879 --> 00:09:34.080 Unnaturally. And here's the kicker. It costs 00:09:34.080 --> 00:09:37.000 $2 for a 5 -second video. Double the price of 00:09:37.000 --> 00:09:40.399 Kling. Exactly. Why pay $2 for a jittery landscape? 00:09:40.740 --> 00:09:43.360 It's just poor value for anything but the most 00:09:43.360 --> 00:09:45.720 controlled slow scenes. And what about the visual 00:09:45.720 --> 00:09:49.860 flaw with WAN AI 2 .6? WAN has great motion logic. 00:09:50.449 --> 00:09:53.690 Falls, jumps, weight shifts, they look genuinely 00:09:53.690 --> 00:09:56.830 believable. But the execution? Is blurry. The 00:09:56.830 --> 00:09:58.789 sources call it the Vaseline on the lens effect. 00:09:58.889 --> 00:10:01.009 It just kills all the fine detail, makes faces 00:10:01.009 --> 00:10:03.909 totally unconvincing. It's cheap, so it's useful 00:10:03.909 --> 00:10:05.850 for testing chordography, but you would never 00:10:05.850 --> 00:10:07.769 use it for a final delivery. And then there's 00:10:07.769 --> 00:10:11.470 Grok Imagine from XAI. The free wildcard. Grok 00:10:11.470 --> 00:10:14.690 is exactly that chaos. It's hilariously imaginative. 00:10:14.889 --> 00:10:17.190 So good for brainstorming. Great for brainstorming 00:10:17.190 --> 00:10:19.490 or meme creation. It gives you 20 free videos 00:10:19.490 --> 00:10:22.710 a day, but it has low detail, robotic audio, 00:10:22.909 --> 00:10:25.429 and the clips are limited to five seconds. So 00:10:25.429 --> 00:10:27.570 definitely not for serious work, but invaluable 00:10:27.570 --> 00:10:30.809 for cost -free experimentation. The overwhelming 00:10:30.809 --> 00:10:33.289 lesson here seems to be avoiding platform loyalty 00:10:33.289 --> 00:10:36.470 at all costs. We've seen a former king get dethroned 00:10:36.470 --> 00:10:39.029 in six months. Rankings change every two weeks. 00:10:39.600 --> 00:10:42.159 Committing to one platform, it just creates this 00:10:42.159 --> 00:10:45.120 sunk cost bias because you've invested time and 00:10:45.120 --> 00:10:47.460 money learning its quirks. You need flexibility 00:10:47.460 --> 00:10:49.980 to protect your budget and your workflow. So 00:10:49.980 --> 00:10:52.620 how do pro users avoid committing to one platform 00:10:52.620 --> 00:10:55.460 if the rankings change that quickly? They use 00:10:55.460 --> 00:10:58.919 aggregators like NVIDIA AI to test one prompt 00:10:58.919 --> 00:11:01.600 across multiple models at the same time. You 00:11:01.600 --> 00:11:04.200 get a real -time comparison. Let's pivot to solutions 00:11:04.200 --> 00:11:05.919 then. We've talked about the problems. We've 00:11:05.919 --> 00:11:08.259 talked about the tools. What are the actionable 00:11:08.259 --> 00:11:10.379 strategies that separate professional control 00:11:10.379 --> 00:11:12.799 from that frustrating guesswork? This is the 00:11:12.799 --> 00:11:15.440 orchestration toolkit. Okay. First, the image 00:11:15.440 --> 00:11:18.320 -to -video consistency hack. This is non -negotiable 00:11:18.320 --> 00:11:21.659 for consistent characters, scenes, items, anything. 00:11:21.940 --> 00:11:24.259 How does it work? Step one, generate your reference 00:11:24.259 --> 00:11:26.200 character in a sophisticated image tool like 00:11:26.200 --> 00:11:29.679 Midjourney. Step two, upload that image to your 00:11:29.679 --> 00:11:33.200 video platform like Kling or VO. Step three. 00:11:33.600 --> 00:11:36.179 animate in image to video mode that is incredibly 00:11:36.179 --> 00:11:38.179 effective but it almost feels like a loophole 00:11:38.179 --> 00:11:40.960 is there a downside does it limit the ai's creativity 00:11:40.960 --> 00:11:43.919 it definitely limits the ai's guesswork which 00:11:43.919 --> 00:11:46.259 is the whole point right text gets interpreted 00:11:46.259 --> 00:11:48.929 differently every single time It creates a visual 00:11:48.929 --> 00:11:51.769 lottery. So think of the image as giving the 00:11:51.769 --> 00:11:55.070 AI a blueprint instead of a suggestion. It locks 00:11:55.070 --> 00:11:57.250 the visual down. So if you want consistency, 00:11:57.470 --> 00:11:59.970 you give it the blueprint. Exactly. You have 00:11:59.970 --> 00:12:01.950 to provide the blueprint. And for professional 00:12:01.950 --> 00:12:04.929 camera work, we use start and end frames for 00:12:04.929 --> 00:12:08.090 that motion control without guessing. Yes. Tools 00:12:08.090 --> 00:12:11.570 like VEO and Kling, they let you define the shot 00:12:11.570 --> 00:12:14.639 precisely. If frame one is a wide shot of the 00:12:14.639 --> 00:12:17.340 character and frame two is a close -up of their 00:12:17.340 --> 00:12:21.419 face. The result is a forced, smooth zoom. A 00:12:21.419 --> 00:12:24.000 smooth zoom or a push -in between those frames. 00:12:24.240 --> 00:12:27.700 You are dictating the camera path. Whoa. Imagine 00:12:27.700 --> 00:12:30.200 scaling this precise frame -by -frame control 00:12:30.200 --> 00:12:32.860 across a feature -length project, not having 00:12:32.860 --> 00:12:34.659 to worry about a camera operator drifting off 00:12:34.659 --> 00:12:36.539 the mark. That is true orchestration. That is 00:12:36.539 --> 00:12:38.779 the power shift. That's it right there. You are 00:12:38.779 --> 00:12:41.750 the director dictating every keyframe. not just 00:12:41.750 --> 00:12:44.289 a describer hoping for a lucky result. And this 00:12:44.289 --> 00:12:46.309 also integrates into our spending strategy. Which 00:12:46.309 --> 00:12:49.250 is? Spend cheap first, spend smart later. Avoid 00:12:49.250 --> 00:12:51.529 burning those expensive premium credits while 00:12:51.529 --> 00:12:54.090 you're just figuring things out. Exactly. Use 00:12:54.090 --> 00:12:57.080 Grok because it's free. Or C -Dance, because 00:12:57.080 --> 00:12:59.539 it's cheap for that whole experimentation phase. 00:12:59.980 --> 00:13:03.100 Dial in your vision, test concepts, iterate cheaply. 00:13:03.559 --> 00:13:06.700 Then, only then, do you generate the final polished 00:13:06.700 --> 00:13:10.000 clips on a premium platform like Kling or VEO. 00:13:10.259 --> 00:13:12.620 And the budget savings are huge. If you follow 00:13:12.620 --> 00:13:15.220 this process, you can reduce your budget by 60 00:13:15.220 --> 00:13:17.620 to 70 percent. We've talked about consistency 00:13:17.620 --> 00:13:20.120 in character and camera, but what about the words 00:13:20.120 --> 00:13:22.700 themselves? Is there a prompt formula that actually 00:13:22.700 --> 00:13:26.409 guarantees quality? Yes. Most prompts fail because 00:13:26.409 --> 00:13:28.750 they're too vague or they focus too much on the 00:13:28.750 --> 00:13:30.450 subject and ignore the environment. So what's 00:13:30.450 --> 00:13:32.570 the formula? The successful formula includes 00:13:32.570 --> 00:13:35.789 five mandatory elements, camera movement, subject 00:13:35.789 --> 00:13:38.350 action, environment, lighting, and specific details. 00:13:38.629 --> 00:13:40.789 So instead of a bad prompt like a woman in a 00:13:40.789 --> 00:13:43.549 lab, what does the good prompt sound like? It 00:13:43.549 --> 00:13:47.019 sounds like this. Medium shot tracking forward, 00:13:47.179 --> 00:13:49.559 a female scientist in a dimly lit laboratory 00:13:49.559 --> 00:13:52.940 carefully examines a glowing specimen. Dramatic 00:13:52.940 --> 00:13:55.480 side lighting casting long shadows, cinematic 00:13:55.480 --> 00:13:58.519 depth of field. So when the AI has those specific 00:13:58.519 --> 00:14:01.919 instructions, those five locked in elements. 00:14:02.120 --> 00:14:04.490 The outputs stop feeling random. And they start 00:14:04.490 --> 00:14:06.610 feeling directed. OK, finally, before we wrap 00:14:06.610 --> 00:14:09.230 up, we need a reality check on the uncanny valley. 00:14:09.470 --> 00:14:11.830 What is the one scenario we should absolutely 00:14:11.830 --> 00:14:14.470 try to avoid with these current gen tools? Avoid 00:14:14.470 --> 00:14:16.870 complex actions like close up hand movements 00:14:16.870 --> 00:14:19.850 while characters are also talking. That combination 00:14:19.850 --> 00:14:22.330 usually breaks both the hands and the lip syncing 00:14:22.330 --> 00:14:26.269 at the same time, which gives you maximum creepiness. 00:14:26.450 --> 00:14:29.190 The big idea for 2026 then is that the best tool 00:14:29.190 --> 00:14:31.429 is simply the one that solves the specific problem 00:14:31.429 --> 00:14:33.860 in front of you. Flexibility and order. orchestration 00:14:33.860 --> 00:14:37.480 beat brand loyalty every single time. So to quickly 00:14:37.480 --> 00:14:39.840 summarize the strategic recommendations by the 00:14:39.840 --> 00:14:42.220 job. If you're making a high -end sci -fi short 00:14:42.220 --> 00:14:44.960 film where quality is everything, you want Kling 00:14:44.960 --> 00:14:49.139 2 .6. For viral TikToks or Reels, you tolerate 00:14:49.139 --> 00:14:52.200 the content restrictions of Sorit 2 because that 00:14:52.200 --> 00:14:55.500 quality is uniquely high for that specific format. 00:14:55.720 --> 00:14:58.519 If you need dialogue, product demos, Anything 00:14:58.519 --> 00:15:00.799 with realistic lip sync and predictable camera 00:15:00.799 --> 00:15:05.539 moves, you go with Google VEO 3 .1. And for serious 00:15:05.539 --> 00:15:08.379 budget projects where you just need volume. Sedans 00:15:08.379 --> 00:15:10.720 1 .5 Pro is the champion there. And the best 00:15:10.720 --> 00:15:13.200 overall value at that combination of price and 00:15:13.200 --> 00:15:15.480 high performance. Sits squarely between Kling 00:15:15.480 --> 00:15:19.820 2 .6 and Sedans 1 .5 Pro. So we've provided the 00:15:19.820 --> 00:15:21.960 ranking and the workflow. Now you have to master 00:15:21.960 --> 00:15:24.279 the execution. Start experimenting with those 00:15:24.279 --> 00:15:27.000 free or cheap tools today. Dial in your vision 00:15:27.000 --> 00:15:28.940 before you commit your premium credits to that 00:15:28.940 --> 00:15:31.279 final generation. And if we agree the core skill 00:15:31.279 --> 00:15:34.019 is now orchestration directing the AI's intelligence, 00:15:34.419 --> 00:15:37.299 we've fundamentally changed the film crew. I 00:15:37.299 --> 00:15:39.799 mean, if the physical camera operator vanishes, 00:15:40.000 --> 00:15:42.679 what unexpected new roles emerge? Maybe prompt 00:15:42.679 --> 00:15:44.620 engineer is just the beginning. Something to 00:15:44.620 --> 00:15:46.659 think about. That's our question for you to consider 00:15:46.659 --> 00:15:49.399 as you dive into this powerful new creative frontier. 00:15:49.679 --> 00:15:51.440 Thank you for joining us for this deep dive. 00:15:51.639 --> 00:15:52.299 Until next time.