← All posts

How I Made a Fern-Style Documentary With AI (No Animation Skills)

Aug 16, 2026
3D documentaryClaudeHiggsfield

Fern-style channels make millions of views per upload with 3D animation so good you forget you’re watching YouTube. I’d never made a 3D animation, with no team and no budget. So I gave myself six hours and made a documentary about the origin of Coca-Cola. Every frame is AI.

The secret: one skill

I spent days reverse-engineering Fern’s channel, including the pacing, the camera moves and how every shot connects, and compressed it into one Claude skill (free in the Vault under YouTube Breakdowns → Fern Style Blueprint). Upload it to Claude and it builds the documentary the way their team would.

Step 1: The production plan

This step decides whether you get Fern or AI slop. I typed: “Make me a Fern-style documentary about the origin of Coca-Cola.” Claude returns a full case file:

  • Asset registry: a locked list of everything that must look identical across the film, like the pharmacist’s face, his coat, the counter and the location. This is where AI documentaries usually fall apart (a beard in scene one, none in scene four).
  • Chapters → scenes, each with narration and a motion prompt (how the camera moves).

Every shot is decided before you generate anything. That’s hours of creative work in one prompt.

Step 2: Assets first

Assets are the anchor. In Higgsfield (50% off) (use desktop), go to Image → Nano Banana Pro and generate every character and location from the registry prompts.

Step 3: Motion in 30-second blocks

Go to Video → Seedance 2.5, attach all your asset images as references, then paste the first scene’s motion prompt.

The trick: I told Claude to build the story in continuous 30-second blocks instead of individual shots. Seedance 2.5 generates the full 30 seconds with up to 50 reference images in 1080p, so the camera moves exactly as planned, with no cuts, no drift and nothing to patch. Repeat for every block.

Step 4: Voice

The narration is already written. In ElevenLabs, use v3 (better emotional control for documentaries), paste it, and click Enhance before generating.

Step 5: Edit

In CapCut or any editor:

  1. Voiceover first. It’s the spine. Strip dead air.
  2. Add clips and match them to the narration, adjusting speed if needed.
  3. Add a quick time/place text overlay at the start with a smooth animation.
  4. Fade in the opening.

Side by side with Fern, the framing, pacing and push-ins feel the same. It’s not identical, since Fern has a team of about 60, but this was one person and one skill.

📝 Your notes on this video

Saved on this device only

Some links in this post are affiliate links. They cost you nothing extra and help support the channel.

New to faceless YouTube?Start with my zero → monetized roadmapNiches, titles, hooks and the monetization rules, all in one place. →

Get my master prompts free

Join the Saidox Faceless Vault: free prompts, checklists and workflows from every video, plus a community of faceless creators.

Join the free Vault →