talking-object

A six-layer recipe for turning a learning objective + an everyday object into a viral-format, educational, Pixar-style talking-object reel. Outputs character design, 5-shot storyboard, image prompts, video prompts, voiceover script, and a self-verification pass — ready to paste into Gemini / Nano Banana and Kling / Veo 3.1 / Higgsfield.

This skill exists because the slop in viral AI video isn't in the models. It's in the brief. A one-shot prompt to Veo produces a man walking past money. A six-layer storyboard, with a frozen character and dialogue lines under their per-shot duration cap, produces something a viewer actually saves.

How to use

1. Tell Claude: "Use the talking-object skill. I want to teach using ."
   • Example: "Use the talking-object skill. I want to teach 'antioxidants protect cells from oxidative stress' using a tomato."
2. On first run, the skill walks you through a one-time config wizard (preferred image tool, preferred video tool, default aspect ratio, voice profile). It saves your config to betterslop-config.yaml in the working directory.
3. The skill returns: a character card, a 5-shot storyboard, 5 image prompts, 5 video prompts, and a voiceover script. All copy-paste ready.
4. You paste into Gemini / Nano Banana to get 5 stills. Paste each still + matching video prompt into Kling / Veo / Higgsfield. Stitch in CapCut.

Layer 1 — Concept lock

The skill refuses to proceed without these three inputs. If the user did not supply them in the trigger, ask.