Running Minimax H3 Cinematic Action Video Workflow on a Single 5060Ti
AppointmentOk972 · reddit · 2026-08-09
The author successfully tested the Minimax H3 Turbo text-to-video model on a single 5060Ti 16GB GPU using ComfyUI.
- Content: Generated a cinematic action scene (agent infiltration) with Japanese dialogue and English subtitles without reference images.
- Workflow: Used larryvrh Lora, 8 steps, sage attention, and INT8 quantization. Generating a single 720p/24fps shot took about 15 minutes.
- Post-processing: Upscaled to 1080p/60fps using Topaz, taking 4.5 minutes.
- Prompting: The highly detailed prompt (camera movement, actions, soundscapes) was generated by Gemini GEM based on a guide.
More from Multimodal
- Meituan LongCat's WorldCast benchmark: GPT-Image-2.5 tops at 72.8, open-source lags — meituan-longcat · 2026-10-16
- Three resources to get up to speed with FLUX 3 — ariG23498 · 2026-10-16
- Watch: Claude Opus 5.5 builds a Tesla Roadster from a single prompt — TAbrodi · 2026-10-16
- Midjourney v8.2 turns out striking indigo ink-wash art with sref 1848359068 — azed_ai · 2026-10-16
- "AI television from another dimension": robot grooms a chihuahua on air — BotLordBotflix · 2026-10-16
- 108M-parameter text-to-image denoiser Surjo-Image-Preview lands on Hugging Face — Consistent-Rule-3720 · 2026-10-16