YouTube2Text

YouTube transcript (eUWlYucdvnI) — Transcript

2,153 words · 235 segments · language en · Watch on YouTube

Full transcript

  1. 0:04Hello, I'm Eunkun Kim.
  2. 0:06This section covers the core design basics for really putting Midjourney to work.
  3. 0:13Have you ever had an experience like this?
  4. 0:15You typed the same word, 'lipstick',
  5. 0:18but sometimes you get a red lipstick, sometimes a pink one,
  6. 0:21some days the lighting is soft,
  7. 0:24and some days the background is dark marble.
  8. 0:26You clearly typed the same words,
  9. 0:29but the result comes out different every time.
  10. 0:31Why is that?
  11. 0:32Many people think of Midjourney as a generator that automatically makes pictures,
  12. 0:37but honestly, that's not enough.
  13. 0:39Midjourney isn't just a tool that outputs images.
  14. 0:43It's a tool for designing visual language.
  15. 0:46Instead of tossing in a few words,
  16. 0:48you need a director's language: lighting, background, material, even mood.
  17. 0:54Without a design, you're leaving the result up to luck.
  18. 0:57But with a structure, you can reproduce the image you want again and again.
  19. 1:02Four goals. First: fully understand the Midjourney web UI, its screens and features.
  20. 1:10Second: the 5-layer prompt structure, designing mood as structured visual language.
  21. 1:18Third: the 4 parameters you'll actually use: version, style, aspect ratio, and seed.
  22. 1:25Fourth: a hands-on mockup exercise, designing three tones for one brand.
  23. 1:34First, let's take a look at the Midjourney web screen.
  24. 1:37When you first open Midjourney,
  25. 1:40you'll see an interface laid out like the one on screen.
  26. 1:43On the left, you'll notice that Explore is checked.
  27. 1:48Explore is active by default when you first log in.
  28. 1:52It's a user gallery page where you can browse images by other members.
  29. 1:57There are various categories along the top.
  30. 2:00You can see a random selection of other users' work.
  31. 2:05Press the Randomize button on the right to reshuffle the list.
  32. 2:13Next is the Hot button.
  33. 2:15As the name says, it shows what's trending right now.
  34. 2:19Next is the Top Day button.
  35. 2:21It shows the daily top list.
  36. 2:23Click the arrow to open the list and switch between Day, Week, and Month.
  37. 2:32Back at the start, look at the top right: Video is currently checked.
  38. 2:38Click Image and you can see the Top Day images as well.
  39. 2:43In Explore you can browse other users' work and collect the pieces you like.
  40. 2:50Works you've hearted like this are gathered under the Likes button.
  41. 2:58You can see the piece I just hearted has been collected there.
  42. 3:02This works for videos too, not just images.
  43. 3:21See the video there as well?
  44. 3:24Using Explore to look at other users' work,
  45. 3:27you can keep up with current AI artwork trends,
  46. 3:31find artwork in the style you're after,
  47. 3:34and check what prompt that user used for reference.
  48. 3:40Next is the For You button.
  49. 3:42As the name suggests, based on your taste,
  50. 3:45these are works Midjourney recommends to you.
  51. 3:51If you're using Midjourney for the first time after logging in,
  52. 3:54a popup like this one will appear.
  53. 3:57That's because Midjourney has no personalization data for a first-time user.
  54. 4:02Building a personalization model is simple.
  55. 4:05Press the orange button and two images appear, left and right.
  56. 4:10Click the one of the two that better matches your taste.
  57. 4:16You need to make 200 picks like this.
  58. 4:18It may feel tedious at first, but once you've personalized,
  59. 4:22Midjourney gets a lot closer to your personal taste.
  60. 4:26Moving on, there's a Styles button at the top right.
  61. 4:30Click it and you'll find models trained on a variety of styles.
  62. 4:34You can check the ones you like to collect them,
  63. 4:38or generate images in the style you want.
  64. 4:47To make an image with Midjourney,
  65. 4:49just type a text prompt into the input field at the top.
  66. 4:53We call this field the prompt box, or the imagine bar.
  67. 4:58Click the settings icon on the right of the input field
  68. 5:00and an options panel opens where you can pick various parameters.
  69. 5:04At the top left is the control for choosing the aspect ratio
  70. 5:08of the image you'll create, and it's very intuitive.
  71. 5:11Slide toward Portrait and the image gets taller;
  72. 5:15slide toward Landscape and it gets wider.
  73. 5:19Hit Reset and it goes back to the default of 1:1.
  74. 5:23Once set, the ratio stays until you change it, even when you run new prompts,
  75. 5:29so the same width-to-height ratio keeps being applied.
  76. 5:33If you type it in the prompt, add --ar 16:9 at the end of the prompt
  77. 5:39and the output comes out in 16:9.
  78. 5:42On the right is the Aesthetics menu.
  79. 5:45Let's go through them one by one.
  80. 5:47Midjourney's model is trained to favor artistic color, composition, and form.
  81. 5:53Stylization controls how strongly that training is applied.
  82. 5:58The default is 100, and you can pick values from 0 to 1000 in steps of 100.
  83. 6:05To set it in the prompt, add --s followed by the value at the end.
  84. 6:11You can adjust it in the options panel or in the text,
  85. 6:16but a parameter typed in the text overrides the value set in the options.
  86. 6:21The same goes for the other parameters.
  87. 6:25Weirdness introduces quirky, unusual qualities into the generated image,
  88. 6:29for when you want unique, unexpected results.
  89. 6:34It's recommended if you want happy accidents, but if you want solid-quality images,
  90. 6:39Midjourney itself advises against using it.
  91. 6:43Personally, when I need visual ideas early in a project,
  92. 6:47I use it now and then.
  93. 6:48The default value is 0.
  94. 6:51In the text, you add --w followed by the value.
  95. 6:56Variety affects how much the initial image grid
  96. 6:58varies from one image to the next.
  97. 7:00The higher the value, the more unusual
  98. 7:02and unexpected the results become.
  99. 7:04Conversely, lower values give you stable,
  100. 7:07repeatable results.
  101. 7:09Simply put, think of it as 'makes odd images'.
  102. 7:14The default value is 0.
  103. 7:17When you want to see a range of looks, adjust this value as you go.
  104. 7:21In the text, you add --c followed by the value.
  105. 7:26These three parameters are what Midjourney calls Aesthetics.
  106. 7:31Stylization, or stylize,
  107. 7:33controls how strongly Midjourney's default aesthetic is applied.
  108. 7:37Weirdness, or weird, sets how unusual an image is versus typical Midjourney output.
  109. 7:45Variety, or chaos, controls how much the initial grid images differ from each other.
  110. 7:51Many people type parameters in the prompt, but I prefer the intuitive options panel
  111. 7:59for adjusting parameters. That covers the basics you'll encounter in Midjourney:
  112. 8:06the overall flow of the interface and the initial settings.
  113. 8:10Once you're comfortable with this screen, you're ready to actually make images.
  114. 8:17Now, what matters is how you type your input.
  115. 8:20Results in Midjourney wobble not because of the UI, but because of prompt structure.
  116. 8:26Remember why the same words give you a different image every time?
  117. 8:31Toss in words with no structure, and the AI interprets them however it likes.
  118. 8:36So now we move to the really important step: making images by design, not by luck.
  119. 8:46Okay, pay close attention from here on. This is the really important part.
  120. 8:50These 5 layers aren't an official Midjourney formula.
  121. 8:53It's a structure directors and creators worldwide settled on for repeatable results.
  122. 9:01It splits the information into units the AI understands best.
  123. 9:05When you write a prompt, instead of throwing words in at random,
  124. 9:08you build it up by stacking these five layers.
  125. 9:13First is Subject: what is the hero of the image?
  126. 9:17For example: lipstick, serum, perfume, white sneakers, coffee cup.
  127. 9:21Nail this down clearly and the AI won't wander.
  128. 9:25Second is Style.
  129. 9:26How will it be expressed?
  130. 9:29Should it look like a photograph,
  131. 9:31like an ad reference, or like artwork? This is where you decide that.
  132. 9:37Product photography, studio shot, editorial beauty, macro shot, cinematic lighting
  133. 9:45Beginners should keep it simple.
  134. 9:48In the example we'll build, 'product photography'
  135. 9:51just that one line makes the image far more stable.
  136. 9:54Third is Technical.
  137. 9:56Technical details: camera, lens, materials, and so on.
  138. 10:00This is the hardest part for beginners, so I'll start with an easy version.
  139. 10:0585mm macro lens, f/2.8, high detail, matte surface, frosted texture, glossy finish
  140. 10:14Change just this and cosmetics, fashion, and packaging images improve instantly.
  141. 10:20Fourth is Environment: background, space, lighting, the world the image lives in.
  142. 10:27White marble, botanical setup, pastel studio, softbox light, rim light, backlit glow
  143. 10:34Get the background and lighting combo right, and the quality is nearly there.
  144. 10:40The fifth and last is Mood.
  145. 10:42This is the finishing step that locks in the brand's emotional tone.
  146. 10:45Luxury minimalism, clean beauty, scientific tone, vegan natural, cinematic soft, elegant modern
  147. 10:53This one line completely changes the image.
  148. 11:05Here's an example using the 5-layer structure.
  149. 11:07As you can see, only the subject changes,
  150. 11:09and the rest of the structure is exactly the same, right?
  151. 11:12Product photography, 85mm lens,
  152. 11:15white marble, soft lighting,
  153. 11:18luxury minimalism,
  154. 11:19--seed 123456,
  155. 11:21--style raw.
  156. 11:22Keep all of that fixed,
  157. 11:24and whether it's a lipstick, a serum,
  158. 11:25a perfume, sneakers,
  159. 11:27or a coffee cup,
  160. 11:28it comes out like one brand campaign,
  161. 11:30with a completely unified tone.
  162. 11:32That's the biggest advantage of the 5-layer structure.
  163. 11:35Swap only the subject and you can batch out images for a brand's entire lineup.
  164. 11:41And with a fixed seed, the light, composition, and tone all stay locked together.
  165. 11:46So in real work, once you build this one structure well,
  166. 11:49you can expand brand artwork really fast.
  167. 11:56Okay, now let's move on to parameters.
  168. 11:59Midjourney has a huge number of options,
  169. 12:01but as a beginner you don't need to know them all.
  170. 12:04Just remember these four.
  171. 12:06First, Version.
  172. 12:08Simply put, it's the engine version.
  173. 12:11Versions range from Niji 4 up to Niji 7,
  174. 12:14and from Version 1 to Version 7.
  175. 12:17Right now, 7 is the most stable and has the strongest quality.
  176. 12:21So I prefer Version 7.
  177. 12:24Second is Style Raw.
  178. 12:26It's the option that suppresses that machine-made AI look as much as possible.
  179. 12:30For product images or brand work especially, consider it almost essential.
  180. 12:35You get a more natural tone, like a real photo shoot.
  181. 12:39Third is Aspect Ratio.
  182. 12:41That's the width-to-height ratio.
  183. 12:43You can keep this really simple.
  184. 12:451:1 for brand images and thumbnails,
  185. 12:4816:9 for regular video,
  186. 12:519:16 for Shorts or vertical video.
  187. 12:53Just these three are more than enough.
  188. 12:56Last is Seed.
  189. 12:58This is the most important one for series work.
  190. 13:00Using the same seed keeps the light, tone, and composition consistent,
  191. 13:04so a brand's full lineup all feels like it was made in the same world.
  192. 13:10Use this one well and your work quality goes up a level.
  193. 13:14So that's the basic design of a prompt.
  194. 13:18Now I'll show you right away how this actually plays out in the images.
  195. 13:28Today we'll use a fictional brand called Aura and create three emotional tones.
  196. 13:34Here's the text prompt example we'll use.
  197. 13:37Lipstick, product photography, 85mm lens, seed, and style raw stay fixed.
  198. 13:44At the end of the prompt, we'll swap between luxury minimalism, vegan natural tone,
  199. 13:48and scientific clean tone, and test each one.
  200. 13:54First, I'll enter the prompt that stays fixed.
  201. 13:57The seed and style raw parameters will stay fixed.
  202. 14:01Let's add 'luxury minimalism' at the end of the prompt.
  203. 14:18Okay, here's the result.
  204. 14:20Using this result as the baseline, let's switch to 'vegan natural tone' and compare.
  205. 14:26On the right of the result, you can see the prompt that was used.
  206. 14:30Below the prompt, the parameters you used appear as buttons.
  207. 14:36Click these buttons to reuse the previous parameters and prompt.
  208. 14:41I'll replace 'luxury minimalism' in the prompt with 'vegan natural tone'.
  209. 14:57We get a result with similar lighting and layout, now with natural elements added.
  210. 15:02You can see the quality of the output stays about the same.
  211. 15:07Next, let's apply 'scientific clean tone' the same way.
  212. 15:13Delete 'vegan natural tone' and type 'scientific clean tone' in its place.
  213. 15:36Done. Since we're using the same seed value, the layout comes out similar too.
  214. 15:42The disappointing part: this last result doesn't feel scientific at all.
  215. 15:48Why? Midjourney weighs the physical elements up front more heavily than mood words.
  216. 15:54'Lipstick, product photography, 85mm lens' already acts like a locked shoot setup,
  217. 16:02so the 'scientific clean tone' at the end can't override it; it just sits on top.
  218. 16:07In other words, changing one word alone doesn't change the tone.
  219. 16:12If you really want a scientific tone, you have to shift the environment itself.
  220. 16:18For example: laboratory background, stainless steel, chrome texture, cold lighting.
  221. 16:24Those elements need to go in first; then the mood words follow naturally.
  222. 16:29That's exactly why the key is recombining the five layers to fit your direction.
  223. 16:36Subject, style, technical, environment, mood: don't just stack them in that order.
  224. 16:44Lead with whichever layer drives your tone most, and it lands far more accurately.
  225. 16:53Once you get this, the classic beginner question, 'I added the word, why no change?'
  226. 16:59is one you can solve right away.
  227. 17:02If you've followed along through Section 5 today,
  228. 17:04you're no longer making images in Midjourney by luck,
  229. 17:06you've reached the stage of making them by design.
  230. 17:09Starting with Section 6,
  231. 17:10we'll use images directly
  232. 17:12to control mood with more precision,
  233. 17:14moving on to the intermediate level.
  234. 17:16Great work today.
  235. 17:18See you in the next section.

About this transcript

This page contains the full transcript of YouTube transcript (eUWlYucdvnI) , generated from the public captions YouTube serves with the video. The transcript has 2,153 words across 235 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.