Running MiniMax H3 locally on a 5070 Ti: multi-shot ad video from a single prompt
Time-Ad-7720 · reddit · 2026-08-20
Inspired by a ComfyUI LinkedIn workflow that used Google and Kling APIs for a one-shot ad, the author challenged themselves to do something more ambitious with only local open-weight models: multiple shots, multiple cuts, all directed through a single prompt.
- Used ComfyUI's MiniMax H3 Reference-to-Video workflow (official docs linked)
- Had ChatGPT structure the video prompt from reference images with per-shot direction (medium close-up opening the can, extreme close-up with can-opening SFX, drinking close-up with gulping SFX, final smile shot) plus SFX and music prompts
- Everything generated locally on an RTX 5070 Ti
Related event: RTX 5070 Ti Runs MiniMax H3 Locally to Generate Multi-Shot Ads(2 posts)→
More from Multimodal
- Open-source anatomy-reveal skill turns Claude into an exploded-view motion design studio — msharmas · 2026-10-16
- Flux 3 Klein vs Flux 2 Klein hands-on: the new base model 'barely usable' for T2I — Both-Rub5248 · 2026-10-16
- USTC paper uses speculative sampling to cut latency in interactive video generation — ustc · 2026-10-16
- 800,000 Gaussian splats run in WebGPU; Casado sees near-infinite scaling — generativist · 2026-10-16
- Nano Banana 2.1 image generation now live in Perplexity Computer for all users — ChrisUniverse · 2026-10-16
- Demo shows precise camera-path control in 3D scenes from a single input image — XRarchitect · 2026-10-16