YouTube2Text

How to Make AI Videos With Claude Fable 5 (Higgsfield Supercomputer) — Transcript

by Parker Prompts · 2,250 words · 321 segments · language en · Watch on YouTube

Full transcript

  1. 0:00Claude Fable 5 just completely changed
  2. 0:02how AI videos get made, and every video
  3. 0:04you're watching right now was made by
  4. 0:05it. Four ads and a week of vlogs, and I
  5. 0:07didn't edit a single frame. I just
  6. 0:09approved steps in one chat. I've been
  7. 0:11testing this setup for the past week,
  8. 0:12and nothing else I've tried even comes
  9. 0:14close. And so, in this video, I'll show
  10. 0:16you exactly how to make AI videos with
  11. 0:18Fable 5 from one brief to a finished
  12. 0:20batch, so you can make your first one
  13. 0:21today. The idea is you stop making the
  14. 0:23videos yourself and start directing an
  15. 0:25agent that makes them for you. With a
  16. 0:27normal video model, you write a prompt
  17. 0:28and take what comes back. But, here you
  18. 0:30hand over the whole job and approve the
  19. 0:31plan, the cast, and each render as it
  20. 0:33builds. Higgsfield supercomputer runs
  21. 0:35all of Higgsfield from a single chat, so
  22. 0:37you describe what you want once, and it
  23. 0:39plans the steps out, matching each one
  24. 0:40to the right model and rendering the
  25. 0:42lot. And the reason it can take on a
  26. 0:43brief this big is that it's running on
  27. 0:45Claude Fable 5. The way it works is that
  28. 0:47Fable 5 is the planner, and each step
  29. 0:50gets handed to whatever specialist fits.
  30. 0:52An image model for the frames, a video
  31. 0:54one for the renders, and a separate
  32. 0:56model for the voices. So, Fable 5's real
  33. 0:58job is lining them all up and passing
  34. 1:00the right thing to each one. Normally,
  35. 1:02four ads like these would mean jumping
  36. 1:03between a few different tools and an
  37. 1:05editor. A week of vlogs on top of that
  38. 1:07is days of work. Here, all of it happens
  39. 1:09in one chat. The only piece I make
  40. 1:11myself is the product, so that's where
  41. 1:12we start. Every ad is going to reuse
  42. 1:14that one product shot, so it's worth
  43. 1:16getting right before anything else.
  44. 1:17Everything today runs on Higgsfield, and
  45. 1:19the link is in the description if you
  46. 1:21want to follow along step-by-step. So,
  47. 1:22before I even open supercomputer, I'll
  48. 1:24go to the image tab, pick GPT image 2,
  49. 1:27set it to a one-to-one square, and
  50. 1:28describe the product I want. For this
  51. 1:30video, the product is a simple mug, a
  52. 1:32matte cream ceramic one with a little
  53. 1:34line art sun on the front. And I've got
  54. 1:35a plain product shot of the mug on a
  55. 1:37white background, which every ad is
  56. 1:39built around from here on. You make it
  57. 1:40once and reuse it, because if you let
  58. 1:42the agent regenerate the mug for every
  59. 1:44ad, the glaze and the shape come out a
  60. 1:45bit different each time, and your four
  61. 1:47ads stop looking like the same product.
  62. 1:49The lighting on this shot matters more
  63. 1:51than it looks. Keep it flat and with no
  64. 1:52harsh shadows or reflections, because
  65. 1:54the more the lighting is locked into the
  66. 1:56product shot, the harder it is for the
  67. 1:57agent to relight it convincingly in a
  68. 1:59sunny kitchen or a dim reading corner
  69. 2:01later. I use GPT image two for it
  70. 2:03because it holds fine product detail
  71. 2:05well. The shape and the little sun come
  72. 2:06out exactly right, which counts for a
  73. 2:08lot when this one image is about to
  74. 2:09appear in every ad. I keep it one-to-one
  75. 2:11on a plain background, too, since a
  76. 2:13simple product plate is the easiest
  77. 2:14thing for the model to drop into any
  78. 2:16scene later. Now, I'll open
  79. 2:17Supercomputer, and there are only two
  80. 2:19settings to change here. I'll pick
  81. 2:20approve before generation, so after
  82. 2:22every prompt it lays out its plan and
  83. 2:23pauses until I hit run instead of
  84. 2:25spinning on its own. I'll set it to
  85. 2:27Claude Fable 5, since that's the part
  86. 2:29that reads the brief and plans every
  87. 2:30step. And then I hand it a single brief
  88. 2:32that has to come back as four ads that
  89. 2:33all feel different. I'll give it that
  90. 2:35brief in one message. The mug goes in
  91. 2:36first, so it's locked as the product
  92. 2:38from line one. Then I tell it what I
  93. 2:40want. UGC style ads, four of them, each
  94. 2:42a different angle, morning coffee, cozy
  95. 2:44reading, office desk, and a gift idea.
  96. 2:46All casual and honest, like real
  97. 2:48creators shot them. I want them
  98. 2:50vertical, 9:16, so they're ready for
  99. 2:52reels and shorts. And I ask it to walk
  100. 2:54me through each step, so I can sign off
  101. 2:56as it goes. This is why I made the image
  102. 2:58first, instead of just describing the
  103. 2:59mug in the brief. If you only describe
  104. 3:01it, the agent pictures it a bit
  105. 3:02differently every time. An attached
  106. 3:04photo gives it one fixed thing to
  107. 3:06[music] copy, so the mug stays identical
  108. 3:07across all four ads. From that one
  109. 3:09message, it comes back with a full plan.
  110. 3:11Four ads and four different avatars,
  111. 3:13each one worked out down to the hair and
  112. 3:15the styling. Everyone gets its own
  113. 3:17setting and its own script. The casting
  114. 3:18is properly varied. The coffee ad gets a
  115. 3:20late 20s woman with ash blonde bangs in
  116. 3:22a bright kitchen, and the reading one a
  117. 3:24softer brunette on a couch with a throw
  118. 3:26and books. For the office ad, it picks a
  119. 3:2830-something guy with light stubble at a
  120. 3:29lived-in desk. And the gift one a
  121. 3:31bearded guy boxing it up under string
  122. 3:33lights. The scripts are just as
  123. 3:34different. [music] One leans on the mug
  124. 3:35making the morning feel calmer. Another
  125. 3:37plays it as a break from a desk full of
  126. 3:39monitors. And the gift version sells it
  127. 3:41as the present that always works. So, no
  128. 3:42two of them sound the same. It all comes
  129. 3:44down to the brief. I tell it to start,
  130. 3:46and it builds the first avatar, a late
  131. 3:4820s woman in a bright kitchen. Once Once
  132. 3:50it she's good, it puts that frame
  133. 3:51together, her holding the mug with the
  134. 3:53sun clearly on it, and starts rendering
  135. 3:55that first video while it moves on to
  136. 3:56the next creator. That start frame step
  137. 3:58is what keeps each ad consistent. It
  138. 4:00locks the creator's face and the mug
  139. 4:02into one still image first. So, when the
  140. 4:04video renders, it has a fixed target to
  141. 4:06animate instead of reinventing her frame
  142. 4:07by frame. [music] I make it check in at
  143. 4:09each step, mostly to save money. Catch a
  144. 4:11bad avatar right here and it's a
  145. 4:12one-line fix. Miss it and let it run and
  146. 4:15you've paid to render four ads you'll
  147. 4:16just throw out. The second creator is a
  148. 4:18woman with soft brown waves and a
  149. 4:19reading mug. And once I approve her, it
  150. 4:21goes straight into her start frame and
  151. 4:22renders the video off that. Then I tell
  152. 4:24it to build the last two at once instead
  153. 4:26of stopping to check with me on each and
  154. 4:28it works the same way through both. The
  155. 4:29avatar first, then the start frame, then
  156. 4:31the video straight off that frame. So,
  157. 4:33now there are four creators who all look
  158. 4:34different, two women and two men with a
  159. 4:36finished ad each. Then I change
  160. 4:37direction. I tell it I want them looking
  161. 4:39at the camera and talking. The video's
  162. 4:41at 10 seconds and the scripts tighten to
  163. 4:43fit. [music] And it regenerates all four
  164. 4:45as 10-second on-camera videos with
  165. 4:46proper lip sync, rewriting each script
  166. 4:48shorter to fit, and it never once starts
  167. 4:50over. I push it to 10 seconds and
  168. 4:52on-camera on purpose. 10 seconds is long
  169. 4:54enough to make one point and short
  170. 4:55enough that nobody scrolls off before
  171. 4:57it's done. And having them talk to
  172. 4:58camera is what makes it read as a real
  173. 5:00person.
  174. 5:01>> This is the mug I reach for first. It's
  175. 5:03just a simple ceramic thing with a sun
  176. 5:05on it,
  177. 5:07but pouring my coffee into it makes the
  178. 5:09whole morning feel calmer.
  179. 5:11This mug is my perfect reading
  180. 5:13companion.
  181. 5:15Tea in one hand, book in the other.
  182. 5:17The little sun on the side just fits the
  183. 5:20vibe.
  184. 5:21>> My desk used to be all screens. This mug
  185. 5:24didn't change much,
  186. 5:26but having it next to my keyboard makes
  187. 5:28the work day feel less robotic.
  188. 5:31Whenever I don't know what to get
  189. 5:32someone, I get this mug.
  190. 5:35Simple, beautiful.
  191. 5:37Everyone deserves a cup that makes them
  192. 5:39smile.
  193. 5:41>> Watching the four back, it's clearly the
  194. 5:42same mug in everyone, just re-lit for
  195. 5:44each scene. Ads are one thing, but the
  196. 5:46same agent will build you a whole
  197. 5:48content series. Now, let's do something
  198. 5:49completely different, a week of casual
  199. 5:51vlogs. This time I want a brand new
  200. 5:53creator instead of reusing any of the ad
  201. 5:55avatars. So, I have it build her
  202. 5:57character sheet first, full body and
  203. 5:58face, and then each day's video comes
  204. 6:00straight off that sheet, so there's no
  205. 6:02start frame to make this time. It's
  206. 6:03seven vlogs, one for each day of her
  207. 6:05week, 10 seconds each and all on camera.
  208. 6:07Then I'll paste the brief in. It plans
  209. 6:09her out, too, a warm mid-20s woman with
  210. 6:11dark hair and light freckles. Then it
  211. 6:13maps out 7 days for her, a morning
  212. 6:15routine, a gym session, errands around
  213. 6:17town, cooking dinner, then a night out
  214. 6:18with friends, a day outdoors, and a slow
  215. 6:20reset day at home. I'll approve, and it
  216. 6:22builds the character sheet, then
  217. 6:24generates all seven vlogs off that one
  218. 6:26sheet, so it's the same face all week.
  219. 6:27Get the character sheet right, and
  220. 6:29you've basically got a creator you own.
  221. 6:30Once she exists, you can put her in any
  222. 6:32video from here on. So, it's worth
  223. 6:34nudging her face and build until they're
  224. 6:35right before you generate a single day.
  225. 6:37It's also why the vlogs come together
  226. 6:39quicker than the ads did. The ads needed
  227. 6:41a fresh start frame for every video,
  228. 6:43since each one was a different creator
  229. 6:44holding a product. Here it's one person
  230. 6:46all week. So, the character sheet on its
  231. 6:47own is enough for the model to hold her
  232. 6:49face the whole way through. The sheet
  233. 6:51keeps her face locked, but the clothes
  234. 6:52and the settings can still change
  235. 6:54day-to-day. If you want a tighter look,
  236. 6:55put that in the brief, too. This vlog
  237. 6:57style is a different job to the ads.
  238. 6:59It's how you'd run a personal brand or a
  239. 7:01faceless channel. I do want to fix the
  240. 7:02voice, though, so I ask it to put a
  241. 7:04single consistent voice across all seven
  242. 7:06days. It offers me a few presets that
  243. 7:08suit a young creator. I pick one called
  244. 7:10Quinn, and it applies that same voice to
  245. 7:12every vlog. Preset voices like that are
  246. 7:14quick, and they fit, but they can sound
  247. 7:15a little generic. So, for a real
  248. 7:17channel, you'd probably clone your own
  249. 7:18voice instead. This is just the fast way
  250. 7:20to keep the whole week sounding like one
  251. 7:22person.
  252. 7:25>> Morning coffee.
  253. 7:28Let's see if I can be a functioning
  254. 7:29human today. Same walk every day, and I
  255. 7:31still love it.
  256. 7:33Three stops, and I've already lost the
  257. 7:35list. Post office, groceries, and one
  258. 7:39thing I definitely forgot.
  259. 7:42Wish me luck.
  260. 7:44Only got like 40 minutes today, so we're
  261. 7:46keeping it short.
  262. 7:48Legs,
  263. 7:49then I'm out.
  264. 7:50Honestly, the hardest part is just
  265. 7:52showing up.
  266. 7:58Attempting to cook something that isn't
  267. 8:01pasta. Wish me luck. Said I'd be ready
  268. 8:04at 8:00. It's 8:15.
  269. 8:06Meeting the girls in 20 minutes.
  270. 8:10This is 5:00 now.
  271. 8:15Needed some sun.
  272. 8:17Found a spot and honestly, might just
  273. 8:20stay here forever. Doing absolutely
  274. 8:23nothing today and I've never felt
  275. 8:25better.
  276. 8:26Laundry's going, tea's made. That's the
  277. 8:29whole plan.
  278. 8:30Reset day.
  279. 8:32>> That's a full week of content built from
  280. 8:33one character sheet and one voice, with
  281. 8:35me doing nothing but describing it. Once
  282. 8:37you've got a recipe that works, you
  283. 8:38don't have to baby sit it forever. You
  284. 8:40can take it off confirm and let it run a
  285. 8:42bigger batch unattended, which is how a
  286. 8:44week of vlogs turns into a month without
  287. 8:46you approving every frame. Because it
  288. 8:47remembers between sessions, the mug, the
  289. 8:49creators and the voice all carry over to
  290. 8:51your next chat. So you're not rebuilding
  291. 8:53your product or your cast every time you
  292. 8:55come back. For anyone making content at
  293. 8:57any volume, this is the change that
  294. 8:59matters. One person can now brief and
  295. 9:00approve what used to take a small team
  296. 9:02and the hard part stops being the
  297. 9:03production and becomes coming up with
  298. 9:05the ideas. The whole batch, all 11
  299. 9:07videos, cost a fraction of what a single
  300. 9:09shoot or an agency day would, because
  301. 9:11you're paying in credits per render
  302. 9:13rather than day rates and revisions. It
  303. 9:15suits anyone who needs volume and
  304. 9:16consistency more than one cinematic hero
  305. 9:18video, a brand running UGC or a creator
  306. 9:20posting daily. And if you only ever
  307. 9:22wanted one perfect 30-second film, you'd
  308. 9:24still art direct that yourself. But for
  309. 9:26a steady stream of content, this is the
  310. 9:27fastest route I've found. So you hand
  311. 9:29Claude Fable 5 and the supercomputer
  312. 9:31your own product or your own week and it
  313. 9:33builds the batch while you approve each
  314. 9:35step. And the part I can't do for you is
  315. 9:37pointed at your own product. [music]
  316. 9:38What it builds from a brief of your own
  317. 9:39is the part I want you to see. Open
  318. 9:41Higgsfield through the first link in the
  319. 9:42description and hand it your first brief
  320. 9:44today. Thank you for watching and I'll
  321. 9:46see you in the next one.

About this transcript

This page contains the full transcript of How to Make AI Videos With Claude Fable 5 (Higgsfield Supercomputer) by Parker Prompts, generated from the public captions YouTube serves with the video. The transcript has 2,250 words across 321 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.