Building a Dragon Ball Z Scene with AI Video Generation | Vilva Demo

Описание: Ever wondered how to turn a single image into an AI-generated Dragon Ball Z episode? In this demo, I walk through the full creative workflow — from generating reference images to creating animated video shots — all using a node-based canvas.
Starting with a Grand Canyon photo, I use Grok and other AI models to generate Goku scenes, iterate on prompts, and connect shots using start/end frames for continuity. You'll see the power of non-linear editing: pick any frame, retry any shot, and build your story one node at a time.
🔗 Try Vilva: https://vilva.ai

What you'll learn:

How to use reference images and AI prompts to generate scenes
Generating videos with models like Veo 3.1
Using start frames and end frames for shot continuity
Extracting frames from generated videos to continue the story
Why node-based workflows beat traditional linear timelines
Tips on prompt engineering for consistent AI video results


Timestamps
0:00 — Intro: Recreating a Dragon Ball Z episode from the Grand Canyon
0:06 — Saving the reference image & generating a prompt with Grok
0:24 — Creating a Goku scene at the Grand Canyon
1:04 — Adjusting aspect ratios for better results
1:18 — Comparing model outputs (Grok vs others)
1:52 — Reviewing generated images
2:08 — Picking reference images & setting 16:9 for video
2:38 — Generating the first video shot (6-second clip)
2:52 — Extracting frames for shot continuity
3:20 — First video result
3:37 — Iterating: trying different models (Veo)
4:29 — Improving hyper-realistic Goku design
5:08 — Better starting point — animating with connected nodes
5:22 — Setting start frame & end frame for Veo 3.1
5:47 — Why node-based systems beat linear editing
6:21 — Continuing from a good frame instead of starting over
6:33 — Reusing the final fight scene prompt
7:00 — Reviewing video quality & critiquing the output
7:25 — Building the Frieza scene: frame extraction demo
7:43 — How frame extraction creates new nodes automatically
8:12 — Connecting extracted frames to new video nodes
8:39 — The importance of prompt engineering
8:48 — Building a library of reusable prompts & workflows
9:17 — Generation time estimates (images: 20s–1min, videos: 2–5min)
9:36 — Final video result
9:47 — Recap: structuring workflows, iterating, and next steps
10:12 — Call to action: try it & share feedback