Summary
This tutorial shows how to produce YouTube Shorts entirely with AI, using the Higgsfield platform as the main toolset. It begins with idea selection, using Claude to research which AI short formats are performing well and which became risky after YouTube changed its Shorts feed on October 1, when original content gained reach and channels re-uploading others’ clips were pushed down.
The chosen concept is a talking animal duo, a cat named Duchess and a dog named Biscuit, arguing about everyday events. To keep the characters consistent across episodes, the creator builds a three-angle character sheet with GPT Image, saves each as a reusable element, and tags them in later prompts.
Voices are generated with ElevenLabs V3 using emotion tags such as bored, smug, and excited. The episode itself is a single 30-second, 9:16, 1080p shot made with Seedance, using the character sheets, a podcast studio image, and voice files as references. A detailed second-by-second prompt controls who speaks and when, and keeps the animals’ mouth movement realistic.
The creator candidly notes the limits of the result: with one static shot and no cuts, it does not feel like a real conversation. The video is a useful look at character consistency, reference-based video generation, and YouTube’s rules on AI content.
📺 Source: Youri van Hofwegen · Published October 29, 2026
🏷️ Format: Tutorial Demo
