AND HOW DOES THAT MAKE YOU FEEL? | An AI Short Comedy Film Made by Claude in Minimax H3 and my Video builder in ComfyUI.
retained text
YouTube Link in case it's still pending [https://youtu.be/TWXF95YT7W8](https://youtu.be/TWXF95YT7W8) 🎬 How this was made My part • One-message brief: a 3-minute comedy in a therapist's office with funny, unique characters and one male doctor: a crying woman, a woman screaming, crying and laughing all at once, a large guy and a skinny old man • Characters made with Z-Image as 3-panel reference sheets • MiniMax H3 2-pass workflow: the Singularity model, with the 8-step LoRA on the second pass at 4 steps and 0.35 denoise • Then I left. I made one call along the way, on a scene that wouldn't behave. What Claude did on its own • Wrote the story, the five characters, all the dialogue and a 25-scene screenplay • Made the cast and the two sets with Z-Image and picked the best seeds • Kept each character's voice description word for word in every scene so the H3 voices stay consistent • Rendered 25 scenes with H3's built-in voices and sound, about 6.3 hours of rendering • QA'd every take: Whisper against the script, pitch and timbre per character, eyelines, frame review sheets • Re-shot the takes that failed: • Scored it with MiniMax Music 3 and screened the cues for accidental vocals • Edited, mastered to -14 LUFS, and checked audio sync on every scene (worst offset 5 ms) 🛠 Tools ComfyUI, VRGDG Video Builder, MiniMax H3 (Singularity ref2va v1.3 + 8-step 768p turbo LoRA), Z-Image Turbo, MiniMax Music 3, Whisper, Claude Code VRGDG nodes: [https://github.com/vrgamegirl19/comfyui-vrgamedevgirl](https://github.com/vrgamegirl19/comfyui-vrgamedevgirl) Go [HERE ](https://youtu.be/uFmSvmv1d5k?si=OTD9ZFQwtc0ABWaF)To watch full walkthrough on how to make video's like this.
document fetched 2026-10-07 00:31 · canonical https://reddit.com/r/StableDiffusion/comments/1wz82c3/and_how_does_that_make_you_feel_an_ai_short