Manhattan at 4 AM: Video Generated with Minimax H3 Tested
cocktailpeanut · x · 2026-08-27
A creator shared a video generated using the Minimax H3 model, depicting "Manhole Runners" in Manhattan at 4 AM.
- Workflow: Used Minimax H3 within WanGP running on Pinokio.
- Comparison: Initially planned to use LTX 2.5 for lip-sync, but found Minimax performed better in all aspects.
- Technical details: Mentioned using FL2VA First Block Cache 30 steps.
More from Multimodal
- Prompt comparison test: Claude vs. ChatGPT vs. Figma Make outputs — Tegadesigns · 2026-08-27
- Adobe Firefly Launches Text-to-Speech and Music Generation — Mr_AllenT · 2026-08-27
- ComfyUI node supports Alibaba's new MiniMax H3 Acc LoRAs — Extension-Yard1918 · 2026-08-27
- AI Music Video: Vacio - Ethics (Official) — Sufficient_Flow_415 · 2026-08-27
- First AI Short Film Made with MiniMax H3: Workflow Recap — GoDavidGo · 2026-08-27
- Google releases Gemini 3.5 Transcribe with 85+ languages and real-time streaming — OfficialLoganK · 2026-08-27