Rookie run: 84-shot video pipeline using rented GPU and MiniMax H3
Legitimate_Bit2775 · reddit · 2026-08-29
A beginner shares a detailed pipeline for generating 84 video shots using MiniMax H3 (Ref2VA) on a rented Vast.ai RTX 4080 Super.
Stack & Config:
- Prep: Claude for story breakdown, ChatGPT for character refs.
- Generation: MiniMax H3 Ref2VA via ComfyUI API, using int8convrot and SageAttention, 6-10 steps (3 min/shot).
- Cost: $0.20/hr.
The author scripted batch processing via the ComfyUI API. While the final cut suffered from narrative issues (too much narration), the experiment validated a low-cost workflow, with plans to focus on scriptwriting next.
More from Multimodal
- Kling 4.0 hands-on: native 30-second 1080p generation and Omni Reference tested — HeyZoyaKhan · 2026-10-15
- Kling 4.0 generates a cinematic Tarzan riding a giant wolf, hard to tell it's AI — HeyZoyaKhan · 2026-10-15
- Kling 4.0 pitches director-style AI video: choreographed action and world-building, not random clips — alex_verem · 2026-10-15
- Adding motion graphics with just a prompt: the Claude Motion x ElevenLabs MCP workflow — CurieuxExplorer · 2026-10-15
- Google turns dense academic PDFs into animated storybooks with Gemini, Nano Banana and Veo — CurieuxExplorer · 2026-10-15
- One product photo plus one Seedance 2.5 prompt yields a full 15-second commercial — CurieuxExplorer · 2026-10-15