A user on Reddit shared a configuration for the MiniMax Turbo Lora (LIGHTX2V) model, which improves the Ref2V audio functionality. The setup involves using the lightx2v turbo lora with the MiniMax-H3 turbo package's TurboSampler node, a sigma shift of 12/3, and 6 steps. This configuration allows for the generation of video from a single reference image and a reference audio sample that differs from the video content. The user demonstrated this by generating a 1-minute, 5-second video in 181 seconds on their system with a 4070 ti GPU and 64GB of RAM, without using EasyCache or Spectrum Apply nodes. AI
IMPACT Provides a specific configuration for users of the MiniMax Turbo Lora (LIGHTX2V) to improve Ref2V audio generation capabilities.
RANK_REASON User-shared configuration for a specific Lora model.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
