Using ChatGPT images as a text-to-motion compiler for AI video with MiniMax H3

cocktailpeanut · x · 2026-11-06

cocktailpeanut demos a new workflow: generate a 15-second animated GIF with ChatGPT (GPT-Image 2.6), then feed it to MiniMax H3 (via Maestro) to produce a full 15-second video. The key insight: ChatGPT images can act as a text-to-motion compiler — no Blender, no manual keyframes, no mocap reference needed. Iterate multiple motion versions via conversation branches, pick the best, and send it to a video generator. Conversation branches replace .blend version files as the versioning mechanism.

Related event: Dev chains GPT-Image GIFs into MiniMax videos, argues creation is shifting to generate-and-select(2 posts)→

Original post →

More from Multimodal

• iPhone-only SpatialAI 3DGS captures rival a professional capture rig, dev says 'solved' — risingodegua · 2026-11-06
• AI-generated Welcome to the NHK game demo shows off anime-grade video quality — HailLordBelial · 2026-11-06
• Where did Z-Image go? Community asks after a year of silence since Base — PixelLunarJelly · 2026-11-06
• AI-generated portrait 'Shadow Play' shows photorealistic skin and lighting detail — aziz4ai · 2026-11-06
• Land Rover Defender 110 concept brand campaign made entirely with Firefly and AI video — ENSAYME · 2026-11-06
• Matthew Berman demos a 4B AI model that generates 3D game assets — Matthew Berman · 2026-11-06

Multimodal channel →