This AI Builds Vox-Style Videos (Multi-lingual) From a Single Sentence — No After Effects — Transcript

Full transcript

1. 0:00I made this entire Vox style explainer
2. 0:02video, the script, the voiceover, every
3. 0:05animated scene from a single sentence.
4. 0:08No After Effects, no motion designer, no
5. 0:11footage, just Claude and one skill I
6. 0:13built. Let me show you. A few years ago,
7. 0:16a video that looks like this needed a
8. 0:18writer, an illustrator, a motion
9. 0:20designer, and a voice actor, a whole
10. 0:23team and days inside After Effects.
11. 0:25Today, it comes down to one
12. 0:26conversation, and I didn't storyboard
13. 0:28it, I didn't prompt every scene. I
14. 0:30didn't even write the script. The skill
15. 0:32did. Here's what everyone gets wrong
16. 0:34about AI motion graphics. Most people
17. 0:37generate one scene at a time, one
18. 0:39prompt, one image, repeat, and every
19. 0:41scene comes out looking slightly
20. 0:43different. Different colors, different
21. 0:45style, but a real Vox style explainer
22. 0:47keeps one visual language from the first
23. 0:49frame to the last. That consistency is
24. 0:52the hard part, and it's exactly what the
25. 0:54skill solves. So, in this video, I'm
26. 0:56going to show you the whole pipeline,
27. 0:58script, voiceover, scene-by-scene
28. 1:00visuals, and final assembly running
29. 1:02inside one chat powered by Higgsfield
30. 1:05through Claude, from one sentence to a
31. 1:06finished video. Let's build it. Before
32. 1:09any of the magic, there's exactly one
33. 1:11manual step, and it takes under a
34. 1:13minute. We connect Claude to Higgsfield,
35. 1:16head to Higgsfield, open MCP and CLI,
36. 1:18and copy your connector link. Back in
37. 1:21Claude, open settings, connectors, add a
38. 1:23custom connector, paste it, name it
39. 1:25Higgsfield, and save. That's it. You
40. 1:27never set it up again.
41. 1:29Every image, every voice, every scene
42. 1:31now flows straight through this
43. 1:32connection. I load the skill, and here's
44. 1:35everything I type. I want a 2-minute 16
45. 1:38by 9 explainer video. You pick the
46. 1:40topic, something people are curious
47. 1:42about right now. Do a conversation with
48. 1:44me. That is the entire brief. I didn't
49. 1:47give it a topic, I didn't give it a
50. 1:48script. I told it how long. That's all.
51. 1:52Watch what it does. It doesn't start
52. 1:53generating, it starts a conversation.
53. 1:56First, it researches what's actually
54. 1:59trending right now and fact-checks the
55. 2:01story against real sources. Then, it
56. 2:03comes back with topic options and asks
57. 2:06about style. I pick a topic, a
58. 2:08flesh-eating parasite crossing the US
59. 2:11border, and the Vox mixed-media collage
60. 2:13look. 30 seconds of decisions,
61. 2:16everything after this happens without
62. 2:18me. Then, it writes the script, not a
63. 2:20blob of text, a structure. 12 blocks,
64. 2:22each one a single spoken line, each one
65. 2:25a scene. Cold open mechanism, history,
66. 2:27the turn, and every factual claim gets
67. 2:29cross-checked against two independent
68. 2:31sources before a word is spoken. 12
69. 2:33facts verified. Now, it runs the
70. 2:35pipeline. It picks one voice and keeps
71. 2:38it for the whole film. It locks a single
72. 2:40style key, one collage look, and
73. 2:42attaches it to every scene. So, scene
74. 2:44one and scene 12 look like the same
75. 2:46designer made them. Then, it generates
76. 2:48the images, the animated clips, the
77. 2:50voice-over, all through Higgsfield, all
78. 2:52in this one chat. And yes, some
79. 2:54generations get rejected and retried.
80. 2:57That's real, I left it in.
81. 2:59One final question, assemble the
82. 3:00project. I say yes, it stitches all 12
83. 3:03clips against the narration, burns the
84. 3:05subtitles, and exports one finished MP4,
85. 3:082 minutes, 16 by 9, done. Total spend
86. 3:11for the whole thing, under 400 credits.
87. 3:13So, here's what one sentence bought.
88. 3:15Every scene, the same visual language.
89. 3:18That is the thing you cannot fake by
90. 3:20generating one image at a time.
91. 3:22>> In June, a 3-week-old calf in Texas was
92. 3:25found with larvae burrowing into its
93. 3:27navel, the first screwworm case on
94. 3:29American soil in 60 years.
95. 3:33The New World screwworm is not a worm.
96. 3:36It is a fly.
97. 3:38And its young will only eat flesh that
98. 3:40is still alive.
99. 3:43A female finds a wound the size of a
100. 3:44tick bite, lays 300 eggs, and within
101. 3:47hours the larvae begin chewing inward.
102. 3:52Ordinary maggots clean tissue. These do
103. 3:55the opposite, drilling deeper as they
104. 3:56feed. And an untreated animal can be
105. 3:58dead within 2 weeks.
106. 4:02America knew this well. Screwworms cost
107. 4:05Southwestern ranchers up to $100 million
108. 4:07a year until 1966 when the country
109. 4:10erased them.
110. 4:12The weapon was absurd. Scientists bred
111. 4:15flies by the billion, sterilized the
112. 4:18males with gamma radiation, and dumped
113. 4:21them out of airplanes. It works because
114. 4:23a female mates exactly once in her short
115. 4:25life. So, one sterile partner ends her
116. 4:28entire bloodline before it starts.
117. 4:32Country by country, the sterile flies
118. 4:35drove the parasites south.
119. 4:37Mexico was clear by 1991.
120. 4:40Panama by 2006. The border holding it
121. 4:43there is not a wall.
122. 4:46It is a factory in Panama breeding 100
123. 4:49million sterile flies every week. Then
124. 4:53the line slipped.
125. 4:54Infestations climbed north through
126. 4:56Central America and Mexico
127. 4:59until Washington finally shut the
128. 5:00southern border to all live cattle. Here
129. 5:02is the strange part. Despite the
130. 5:04headlines, not one person in the United
131. 5:07States has caught it.
132. 5:09Locally,
133. 5:10this is a war over livestock.
134. 5:13That calf was not bad luck. It was a
135. 5:15warning. And the answer has not changed
136. 5:17since 1966. Breed them faster than they
137. 5:19breed.
138. 5:22>> Go through it scene by scene and it
139. 5:24feels like one production because it is.
140. 5:27The style is locked once and applied
141. 5:29everywhere. And this isn't only for Vox
142. 5:31Style. The same skill does
143. 5:33documentaries, history, finance,
144. 5:35science. The workflow never changes.
145. 5:38Think about where this goes. Faceless
146. 5:40YouTube channels in finance, history,
147. 5:43geopolitics, brand explainers, and
148. 5:45product walkthroughs, marketing videos,
149. 5:48educational content, one creator, one
150. 5:50conversation, the whole pipeline. But,
151. 5:53here's the part that actually got me.
152. 5:55After the video was done, I typed one
153. 5:57more line, "Now, make this in Hindi,
154. 5:59same visuals, new voice over." A few
155. 6:01minutes later, a second complete video.
156. 6:03You're not translating subtitles, you're
157. 6:05shipping a native version into every
158. 6:07market for almost nothing.
159. 7:11This is the shift. You don't learn the
160. 7:13pipeline anymore. You describe the
161. 7:15result, and the pipeline already lives
162. 7:17inside the skill. That's the whole
163. 7:19point, one sentence and this video. The
164. 7:21distance between those two things is
165. 7:23everything. You can try Higgs Field for
166. 7:26free, the link is in the description,
167. 7:28and the MCP setup takes less than a
168. 7:30minute. I'm dropping the skill and the
169. 7:32full workflow down there, too. So, go
170. 7:34make your first Vox style video, and
171. 7:36I'll see you in the next one.