WEBVTT 00:00:00.000 --> 00:00:03.720 creating AI art is. It's genuinely fun, isn't 00:00:03.720 --> 00:00:07.080 it? You dial in the exact hair, the precise texture 00:00:07.080 --> 00:00:09.179 of a hoodie, that perfect expression, and you 00:00:09.179 --> 00:00:11.400 get this feeling that you've, you know, you've 00:00:11.400 --> 00:00:13.460 totally mastered the machine. You feel like a 00:00:13.460 --> 00:00:16.710 director who just nailed the casting. But ah, 00:00:16.829 --> 00:00:19.370 that feeling is often pretty fleeting. Exactly. 00:00:19.809 --> 00:00:21.670 Then you want to move that perfect character 00:00:21.670 --> 00:00:24.449 to a new background, or maybe just change their 00:00:24.449 --> 00:00:27.250 gaze, and the AI serves you a complete stranger. 00:00:27.850 --> 00:00:29.649 Suddenly the face geometry is different, the 00:00:29.649 --> 00:00:31.910 hoodie color has shifted, and you're back to 00:00:31.910 --> 00:00:35.009 playing this frustrating AI lottery. It's exhausting. 00:00:35.710 --> 00:00:37.329 Trying to describe what you already made with 00:00:37.329 --> 00:00:41.049 these complex 50 -word text prompts. It's a huge 00:00:41.049 --> 00:00:43.250 creative bottleneck. This is where we break that 00:00:43.250 --> 00:00:46.579 cycle. Welcome back to the deep dive. Our mission 00:00:46.579 --> 00:00:48.960 today is to unpack a tool that really changes 00:00:48.960 --> 00:00:52.340 the physics of content creation, Google Whisk. 00:00:52.600 --> 00:00:55.219 This is Google's dedicated solution for consistency, 00:00:55.659 --> 00:00:57.579 and it's powered by their most advanced image 00:00:57.579 --> 00:01:01.159 model, Imagen 3. The goal here is simple, giving 00:01:01.159 --> 00:01:03.159 you professional -grade control. We're going 00:01:03.159 --> 00:01:05.000 to walk you through the system's core genius. 00:01:05.319 --> 00:01:07.620 It separates the subject, the scene, and the 00:01:07.620 --> 00:01:10.480 style. It lets you anchor your character like 00:01:10.480 --> 00:01:12.780 a physical pin on a board. And we'll show you 00:01:12.780 --> 00:01:15.480 why this is probably the only way to really scale 00:01:15.480 --> 00:01:18.480 your AI art, plus cover some advanced tips for 00:01:18.480 --> 00:01:21.159 refining and editing your final images. OK, so 00:01:21.159 --> 00:01:23.140 let's start with the root of this frustration. 00:01:23.840 --> 00:01:26.540 Why does this AI lottery even exist in the first 00:01:26.540 --> 00:01:29.400 place? Well, because when you use generators 00:01:29.400 --> 00:01:32.959 like Mid Journey or Delhi, every single time 00:01:32.959 --> 00:01:36.359 you hit generate, the AI is starting over. From 00:01:36.359 --> 00:01:38.420 scratch. Completely from scratch. It's just drawing 00:01:38.420 --> 00:01:40.579 a new picture based on the words you gave it. 00:01:40.670 --> 00:01:44.069 It retains zero visual memory of the last image. 00:01:44.290 --> 00:01:46.849 So even with an identical prompt, the randomness 00:01:46.849 --> 00:01:49.650 is different, and that leads to that visual drift 00:01:49.650 --> 00:01:51.790 in the character's face. It's like asking a different 00:01:51.790 --> 00:01:54.109 artist to redraw the same person every single 00:01:54.109 --> 00:01:55.969 time, but they only have a written description 00:01:55.969 --> 00:01:58.549 to go on. And Google Whisk just throws that whole 00:01:58.549 --> 00:02:01.189 playbook out. Instead of just relying on long, 00:02:01.290 --> 00:02:04.189 wordy descriptions, it uses images as the primary 00:02:04.189 --> 00:02:07.340 guide. Right. You generate a character, you drag 00:02:07.340 --> 00:02:09.780 the photo into the sidebar, and you lock it. 00:02:10.020 --> 00:02:13.780 That locked visual data is transformative. It 00:02:13.780 --> 00:02:16.759 allows the AI to remember the exact facial geometry, 00:02:17.219 --> 00:02:19.080 the specific hair color, the precise shape of 00:02:19.080 --> 00:02:21.719 the body. Which, if you're creating a comic book, 00:02:21.960 --> 00:02:25.240 or a branded avatar, or really any story where 00:02:25.240 --> 00:02:28.439 recognition matters, that becomes indispensable. 00:02:28.659 --> 00:02:30.560 And what's fascinating is the tech behind it. 00:02:30.759 --> 00:02:33.599 Whisk is built on Imogen 3, which is Google's 00:02:33.599 --> 00:02:35.960 top -tier image model right now. And Imogen 3 00:02:35.960 --> 00:02:39.759 is really good at understanding small, complex 00:02:39.759 --> 00:02:42.719 requests. So it can maintain that character anchor 00:02:42.719 --> 00:02:45.419 while you make really subtle changes to the scene 00:02:45.419 --> 00:02:48.360 or the action. That anchoring is possible because 00:02:48.360 --> 00:02:51.020 Whisk enforces what they call a separation principle. 00:02:51.500 --> 00:02:53.520 The background is one layer, the subject is another. 00:02:53.719 --> 00:02:56.139 So you can swap people into old scenes or change 00:02:56.139 --> 00:02:58.099 the style without ever distorting the character's 00:02:58.099 --> 00:02:59.659 face. It's really accessible, too. It's just 00:02:59.659 --> 00:03:01.909 drag and drop. So, okay, I'm trying to get my 00:03:01.909 --> 00:03:04.449 head around this. How does locking the image 00:03:04.449 --> 00:03:07.770 actually stop the AI from starting fresh every 00:03:07.770 --> 00:03:10.509 time? It's because it locks that core visual 00:03:10.509 --> 00:03:13.590 reference data. It forces the separation of the 00:03:13.590 --> 00:03:16.389 subject's identity from the scene's data. Ah, 00:03:16.569 --> 00:03:19.099 okay. So it's smart -pursing. It knows what to 00:03:19.099 --> 00:03:21.539 keep and what to change. Exactly. Got it. Now, 00:03:21.539 --> 00:03:23.259 before we start locking down our characters, 00:03:23.819 --> 00:03:25.819 there is a little bit of setup. We have to decide 00:03:25.819 --> 00:03:28.659 on the aspect ratio, the final shape of the canvas. 00:03:29.080 --> 00:03:31.479 And this step is surprisingly critical. People 00:03:31.479 --> 00:03:34.180 overlook it all the time. Really? Oh, yeah. If 00:03:34.180 --> 00:03:36.259 you choose the wrong ratio, you end up cropping 00:03:36.259 --> 00:03:38.599 your main subject later. And that kind of defeats 00:03:38.599 --> 00:03:41.419 the whole purpose of getting consistency. Right. 00:03:41.479 --> 00:03:43.759 So when you get into the WISC interface, you 00:03:43.759 --> 00:03:45.599 look towards the bottom of the screen. There's 00:03:45.599 --> 00:03:48.319 a little icon that lets you pre -select the dimensions. 00:03:48.620 --> 00:03:51.259 So 1 .1 for squares. Perfect for standard social 00:03:51.259 --> 00:03:54.080 media posts. Right. 9 .16 for portrait, which 00:03:54.080 --> 00:03:57.740 is for phones, stories. And then 16 .9 landscape. 00:03:57.840 --> 00:04:00.360 That's your cinematic, your YouTube intro look. 00:04:00.409 --> 00:04:04.009 OK, now for the key. The secret to flawless consistency 00:04:04.009 --> 00:04:06.569 later on is creating the perfect original character 00:04:06.569 --> 00:04:10.370 first. And to avoid errors down the line, that 00:04:10.370 --> 00:04:12.810 initial character has to be created against the 00:04:12.810 --> 00:04:14.969 simplest background possible. This is that green 00:04:14.969 --> 00:04:16.990 screen tip we were talking about. Exactly. You 00:04:16.990 --> 00:04:19.970 need to use phrasing in your prompt, like standing 00:04:19.970 --> 00:04:24.009 against a plain green studio background, or solid 00:04:24.009 --> 00:04:27.310 gray, solid blue, whatever. So why is that plain 00:04:27.310 --> 00:04:30.750 background step so vital? It seems a little counterintuitive 00:04:30.750 --> 00:04:33.009 when the goal is to put them in a complex scene 00:04:33.009 --> 00:04:35.470 later. Well, think about it like a real movie 00:04:35.470 --> 00:04:38.389 production. OK. You film an actor against a green 00:04:38.389 --> 00:04:41.350 screen. So the editors can cleanly isolate them, 00:04:41.350 --> 00:04:44.009 right? Edge to edge. We're putting them into 00:04:44.009 --> 00:04:47.829 a CGI world. The AI needs that exact same clean 00:04:47.829 --> 00:04:50.310 break. It helps the machine distinguish the person 00:04:50.310 --> 00:04:52.569 from the setting so you don't get messy background 00:04:52.569 --> 00:04:54.610 bits stuck to them. So you're just simplifying 00:04:54.610 --> 00:04:57.259 the AI's job from the get -go. Precisely. You're 00:04:57.259 --> 00:04:59.680 setting it up for success by reducing noise first. 00:04:59.939 --> 00:05:02.620 Makes perfect sense. After all that setup, we 00:05:02.620 --> 00:05:05.459 finally get to the payoff. Looking at the sidebar 00:05:05.459 --> 00:05:07.939 on the screen, you see those three powerful empty 00:05:07.939 --> 00:05:11.819 boxes, subject, scene, and style. We start by 00:05:11.819 --> 00:05:14.319 anchoring the actor. Yeah. The subject. Once 00:05:14.319 --> 00:05:17.120 you've generated that perfect image, let's say 00:05:17.120 --> 00:05:18.879 it's our blonde young man in the blue hoodie, 00:05:19.360 --> 00:05:22.569 you click it. drag it into the subject box, and 00:05:22.569 --> 00:05:24.810 then, this is important, you click the little 00:05:24.810 --> 00:05:27.970 ticker checkbox to activate the lock. That actor 00:05:27.970 --> 00:05:30.509 is now hired and anchored, and the efficiency 00:05:30.509 --> 00:05:33.209 you gain here is just immense. Wait, hold on, 00:05:33.269 --> 00:05:35.810 if I lock the subject, are you saying I can just 00:05:35.810 --> 00:05:38.089 cut out most of my description? I don't have 00:05:38.089 --> 00:05:39.970 to keep writing blonde, boy, and blue hoodie 00:05:39.970 --> 00:05:42.350 for the next hundred images. You never have to 00:05:42.350 --> 00:05:44.870 write it again. You've replaced that description 00:05:44.870 --> 00:05:47.990 with a visual memory, so your prompt only needs 00:05:47.990 --> 00:05:50.750 to describe the new location or the action. That's 00:05:50.750 --> 00:05:53.449 amazing. So the prompt just becomes walking through 00:05:53.449 --> 00:05:56.230 a busy street in Tokyo at night, neon lights 00:05:56.230 --> 00:05:59.189 reflecting on wet pavement, cinematic atmosphere. 00:05:59.430 --> 00:06:01.649 And the result is the same boy, same hoodie, 00:06:01.889 --> 00:06:04.350 same features instantly transplanted into that 00:06:04.350 --> 00:06:06.550 new hyper -detailed environment. It's like hiring 00:06:06.550 --> 00:06:08.449 the same lead actor and just moving them from 00:06:08.449 --> 00:06:11.220 one movie set to another. Exactly. It's a huge 00:06:11.220 --> 00:06:13.480 shortcut for any kind of sequential content. 00:06:14.180 --> 00:06:16.720 Next up, let's look at that second powerful lock. 00:06:17.899 --> 00:06:20.680 The scene. The scene lock. This is for when you 00:06:20.680 --> 00:06:23.420 need to keep the location fixed. The exact room, 00:06:23.639 --> 00:06:26.639 the lighting, the camera angle. All of it. You're 00:06:26.639 --> 00:06:29.300 building a permanent movie set. You are. So you 00:06:29.300 --> 00:06:32.180 start by creating a really detailed space. Something 00:06:32.180 --> 00:06:35.800 like a hyper -realistic interior of a cozy mountain 00:06:35.800 --> 00:06:39.000 cabin at night. Large glass windows showing a 00:06:39.000 --> 00:06:42.620 heavy snowstorm outside. A warm fireplace glowing. 00:06:43.079 --> 00:06:46.079 Hmm, very specific. I can picture it. Once you 00:06:46.079 --> 00:06:48.500 have that perfect cabin image, you drag it into 00:06:48.500 --> 00:06:51.019 the scene box and you lock it. Every new image 00:06:51.019 --> 00:06:52.819 you generate happens right there. Okay, so once 00:06:52.819 --> 00:06:54.879 the scene is set, you can remove the old subject 00:06:54.879 --> 00:06:57.639 image or just leave that box empty and write 00:06:57.639 --> 00:07:00.019 a new prompt focusing on who or what you want 00:07:00.019 --> 00:07:02.860 to place there. So we can write a cute golden 00:07:02.860 --> 00:07:04.939 retriever puppy sitting on the floor of the room. 00:07:05.259 --> 00:07:07.540 And the puppy just appears? It just appears right 00:07:07.540 --> 00:07:09.759 there in the fixed cabin. The walls, the fireplace, 00:07:09.959 --> 00:07:13.180 the window. It's all perfectly unchanged. It's 00:07:13.180 --> 00:07:15.980 how you keep connected stories visually coherent. 00:07:16.220 --> 00:07:18.420 So if both the subject and the scene are locked, 00:07:18.680 --> 00:07:21.839 What's left to customize? What can the text prompt 00:07:21.839 --> 00:07:25.300 still change? Well, primarily specific actions 00:07:25.300 --> 00:07:28.720 that you write in the new prompt. But also, critically, 00:07:29.339 --> 00:07:33.060 style changes. The geometry is set, but the visual 00:07:33.060 --> 00:07:35.500 vibe can still be completely transformed. Which 00:07:35.500 --> 00:07:38.740 brings us to that third pillar, style. The style 00:07:38.740 --> 00:07:42.639 box manages the soul of the picture, the art 00:07:42.639 --> 00:07:45.779 language. Right. Is it a real -life photo? A 00:07:45.779 --> 00:07:48.930 messy watercolor? Maybe that whimsical claymation 00:07:48.930 --> 00:07:50.810 you mentioned. This is where you set the texture 00:07:50.810 --> 00:07:52.209 and the color palette. And you don't have to 00:07:52.209 --> 00:07:54.550 be an art historian. You just find a sample image 00:07:54.550 --> 00:07:56.790 that has the kind of lines or color grading you 00:07:56.790 --> 00:08:00.069 like and drag it into the style box. An effective 00:08:00.069 --> 00:08:03.129 style prompt could be something like whimsical 00:08:03.129 --> 00:08:06.290 claymation style, soft lighting. vibrant pastel 00:08:06.290 --> 00:08:08.490 colors, textured clay surfaces. And that gives 00:08:08.490 --> 00:08:11.189 the AI all the cues it needs for the look? Now 00:08:11.189 --> 00:08:13.629 imagine the power of triple locking. Okay. You 00:08:13.629 --> 00:08:15.790 lock the blonde boy as the subject, the Tokyo 00:08:15.790 --> 00:08:18.189 street as the scene, and this clay nation vibe 00:08:18.189 --> 00:08:21.290 as the style. Whoa. So you'd get a picture of 00:08:21.290 --> 00:08:24.069 a clay Tokyo street with that same recognizable 00:08:24.069 --> 00:08:26.529 clay boy. Consistent features but a totally different 00:08:26.529 --> 00:08:29.629 material. The consistency is total. And that's 00:08:29.629 --> 00:08:31.949 the moment of wonder, right? When you realize 00:08:31.949 --> 00:08:35.340 you can scale this up to produce a full animated 00:08:35.340 --> 00:08:38.220 series, or professional storyboards with perfect 00:08:38.220 --> 00:08:40.779 visual consistency. It changes everything for 00:08:40.779 --> 00:08:43.220 high -volume creators. Absolutely, and you can 00:08:43.220 --> 00:08:45.659 use almost anything to define the style, like 00:08:45.659 --> 00:08:49.039 a high -quality landscape photo. Its color theme 00:08:49.039 --> 00:08:51.340 can define the style for a close -up portrait. 00:08:51.620 --> 00:08:54.039 So the color grading matches perfectly across 00:08:54.039 --> 00:08:57.159 different kinds of images? It's pure visual cross 00:08:57.159 --> 00:08:59.159 -referencing. Okay, so once you have your perfectly 00:08:59.159 --> 00:09:02.379 consistent image, Google WISC gives you two final 00:09:02.379 --> 00:09:05.870 options for fine -tuning. and refine. And they 00:09:05.870 --> 00:09:08.370 serve very different functions. Knowing when 00:09:08.370 --> 00:09:11.049 to use which one is really key. So edit is for 00:09:11.049 --> 00:09:13.990 the big stuff. Big changes, replacements. You 00:09:13.990 --> 00:09:15.710 type change the blue hoodie to a red leather 00:09:15.710 --> 00:09:18.250 jacket or add a pair of mirrored sunglasses. 00:09:18.629 --> 00:09:20.990 It replaces elements completely. And refine is 00:09:20.990 --> 00:09:23.889 more subtle. Refine is the final polish, making 00:09:23.889 --> 00:09:26.169 the sky a little brighter, fixing a small glitch, 00:09:26.429 --> 00:09:28.830 ensuring the face clarity is perfect. It's the 00:09:28.830 --> 00:09:31.779 touch up tool. And then there's this huge control 00:09:31.779 --> 00:09:34.580 knob hidden away in the settings menu, the precise 00:09:34.580 --> 00:09:36.700 reference switch. Oh, yeah, this is a big one. 00:09:36.840 --> 00:09:40.139 When that switch is on, the AI follows your subject 00:09:40.139 --> 00:09:44.360 sample with incredible fidelity, near 100 % accuracy. 00:09:44.539 --> 00:09:48.779 It locks down every single detail. But when it's 00:09:48.779 --> 00:09:53.080 O off S, the AI is much more free. It takes inspiration 00:09:53.080 --> 00:09:55.899 from the subject. but it allows for variation. 00:09:56.080 --> 00:09:57.700 Which is what you need if you want your character 00:09:57.700 --> 00:09:59.779 to show different facial expressions. Exactly. 00:09:59.899 --> 00:10:03.399 Fear. or laughter or changing their pose without 00:10:03.399 --> 00:10:06.360 you having to manually define every little muscle 00:10:06.360 --> 00:10:08.659 movement. You know, I'll admit, I still wrestle 00:10:08.659 --> 00:10:11.299 with prompt drift myself when I try to get characters 00:10:11.299 --> 00:10:13.700 to express emotion. Oh, me too. It's a constant 00:10:13.700 --> 00:10:15.820 battle. You're asked for a scowl, you get a smirk. 00:10:16.240 --> 00:10:18.740 So that precise reference switch, that sounds 00:10:18.740 --> 00:10:21.320 incredibly useful for getting that creative variance 00:10:21.320 --> 00:10:23.860 without losing the core identity. Now, here's 00:10:23.860 --> 00:10:26.120 an advanced tip, but it's a critical one. OK. 00:10:26.820 --> 00:10:29.419 Simplicity is key when the sidebar is locked. 00:10:29.519 --> 00:10:32.080 If the subject and scene are already locked visually 00:10:32.080 --> 00:10:34.799 in the sidebar keep your main text prompt simple 00:10:34.799 --> 00:10:38.340 like just like eating ramen or running in the 00:10:38.340 --> 00:10:40.600 rain. Because if you write a long, detailed paragraph 00:10:40.600 --> 00:10:42.779 describing the blonde man in the blue hoodie 00:10:42.779 --> 00:10:45.500 eating ramen. The AI starts getting confused. 00:10:45.720 --> 00:10:47.679 It's trying to balance the perfectly locked visual 00:10:47.679 --> 00:10:50.259 sample with this huge pile of redundant words. 00:10:50.740 --> 00:10:53.019 The AI starts to fight itself, trying to balance 00:10:53.019 --> 00:10:55.980 the locked references with the unnecessary text. 00:10:56.299 --> 00:10:58.340 Keep the prompt focused only on what needs to 00:10:58.340 --> 00:11:01.360 change. Got it. OK, let's quickly synthesize 00:11:01.360 --> 00:11:03.299 the core tools we've covered in this deep dive. 00:11:03.580 --> 00:11:07.039 The subject box. It remembers the face, the body, 00:11:07.299 --> 00:11:10.059 the clothing, and the pro tip is still. Start 00:11:10.059 --> 00:11:13.019 on a plain background. The scene box. Yeah. It 00:11:13.019 --> 00:11:15.200 keeps the room structure, the lighting, the camera 00:11:15.200 --> 00:11:18.000 angle fixed. You are building a repeatable, permanent 00:11:18.000 --> 00:11:20.519 set. Style copies the art language, the materials, 00:11:20.679 --> 00:11:23.200 the texture, the colors, turns a photo into a 00:11:23.200 --> 00:11:25.580 pencil sketch instantly. And finally, edit handles 00:11:25.580 --> 00:11:28.419 the large scale replacements, while refine is 00:11:28.419 --> 00:11:30.850 for the final polish. And this opens up some 00:11:30.850 --> 00:11:33.269 incredibly practical real world applications. 00:11:33.769 --> 00:11:36.590 Take social media. You can create a recognizable 00:11:36.590 --> 00:11:40.250 virtual model for your brand. That model can 00:11:40.250 --> 00:11:43.190 consistently introduce products, travel the world. 00:11:43.570 --> 00:11:46.370 It builds immediate brand recognition through 00:11:46.370 --> 00:11:49.370 a consistent visual identity. And for storytelling. 00:11:49.840 --> 00:11:52.919 This tool is priceless. Storyboarding used to 00:11:52.919 --> 00:11:55.440 be this massive time sink for artists, right? 00:11:55.940 --> 00:11:58.299 Redrawing characters over and over. Frame after 00:11:58.299 --> 00:12:00.340 frame. Now you just anchor the character once 00:12:00.340 --> 00:12:02.039 and you can instantly throw them into different 00:12:02.039 --> 00:12:05.159 camera shots. Imagine locking a character like 00:12:05.159 --> 00:12:08.960 a cybernetic dog as the subject and using a prompt 00:12:08.960 --> 00:12:11.559 like detailed pencil sketch style storyboard 00:12:11.559 --> 00:12:14.000 panel, the dog looking surprised, high angle 00:12:14.000 --> 00:12:16.259 shot. You get a professional comic book frame 00:12:16.259 --> 00:12:19.299 perfectly consistent in seconds. This completely 00:12:19.299 --> 00:12:21.220 changes the economics of visual storytelling. 00:12:21.480 --> 00:12:23.419 So this whole process is really just a smart 00:12:23.419 --> 00:12:25.899 puzzle game based on visual layering. Absolutely. 00:12:25.899 --> 00:12:27.860 Don't be afraid of making mistakes. The AI doesn't 00:12:27.860 --> 00:12:30.100 get tired. And remember to manage your library. 00:12:30.539 --> 00:12:32.539 Always favorite the images that are beautiful 00:12:32.539 --> 00:12:34.960 or useful. And save the seed number if you like 00:12:34.960 --> 00:12:37.279 a particular layout. You can even use the animate 00:12:37.279 --> 00:12:39.659 feature for short moving videos if you have the 00:12:39.659 --> 00:12:42.320 credits. But most importantly, we want to challenge 00:12:42.320 --> 00:12:45.059 you to practice this right away. OK. Go to Google 00:12:45.059 --> 00:12:48.480 Labs. Create a simple character. maybe a quirky 00:12:48.480 --> 00:12:51.220 cat in boots, and use the subject box to lock 00:12:51.220 --> 00:12:53.759 that cat. And then? Then place that anchored 00:12:53.759 --> 00:12:56.860 cat into a hyper detailed outer space background. 00:12:57.480 --> 00:12:59.840 Go solve that puzzle and just feel the power 00:12:59.840 --> 00:13:02.740 of consistent creation. So what does this all 00:13:02.740 --> 00:13:05.220 mean for the bigger picture if this level of 00:13:05.220 --> 00:13:08.259 precise repeatable control now makes professional 00:13:08.259 --> 00:13:11.320 consistent storyboarding nearly instant and accessible 00:13:11.320 --> 00:13:14.679 to anyone? What happens to the demand for human 00:13:14.679 --> 00:13:17.460 conceptual artists in a few years? That's the 00:13:17.460 --> 00:13:19.480 heavy question, isn't it? How do human creators 00:13:19.480 --> 00:13:21.940 shift their focus when the machine can handle 00:13:21.940 --> 00:13:24.279 the grunt work of visual consistency? Something 00:13:24.279 --> 00:13:26.639 for you to mull over until our next deep dive.