🧠 DualSpace MiniMax-H3 — Text/Vision Conditioner Service (I2V)

Space ini berfungsi sebagai Conditioner as a Service (Backend) untuk memproses Qwen3-VL 32B Text/Vision Encoder dan Video VAE Keyframes. Mengembalikan file .safetensors (Thin Wire Protocol) untuk di-stream ke Space Generator.

Durasi Video

🧠 Status Conditioner Service (Space 1 - Hot Standby)

  • PyTorch Version: 2.13.0+cu130
  • CUDA Available: True
  • CUDA Version: 13.0
  • Device Name: NVIDIA RTX PRO 6000 Blackwell Server Edition
  • VRAM: 94.97 GB
  • Compute Capability: 12.0
  • Model Loaded: Qwen3-VL 32B NVFP4/AWQ (~15.7 GB)
  • Model VAE: minimax_h3_video_vae_int8_convrot (~2.6 GB)
  • Mode: Image-to-Video (I2V) Hot-Standby Pre-Warmed Engine
  • Dynamic Duration: 20s (5s) | 25s (10s/15s)
  • Custom Nodes: 0 External Repos (Pure ComfyUI Core Native)
  • Wire Output: .safetensors binary format