My approach: I generate everything in "Fast" mode first to quickly iterate and find the best results, then I selectively upscale only the videos worth keeping by reusing the same seed.<h2 id="generation-modalities">Generation Modalities</h2><ul><li>📝➡️🎥 Text-to-Video Create completely new videos from scratch using text prompts.</li><li>🖼️➡️🎥 Image-to-Video Animate static reference images using text prompts.</li><li>🖼️🖼️➡️🎥 First-Last-Frame-to-Video Generates a coherent video sequence that bridges a defined starting and ending image.</li><li>🎥➡️🎥 Video-to-Video Generates synchronized audio and speech driven by video visuals. Offers options to pass through or entirely regenerate the source video.</li></ul><h2 id="audio-input-settings">Audio Input Settings</h2><ul><li>🔇 No Audio Input No external audio file used. The AI generates new audio based on your text prompt.</li><li>🔊 Audio Input Upload an existing voice or music file to drive the animation and lip-sync.</li></ul><h2 id="vram-optimization-and-long-videos">🚀 VRAM Optimization &amp; Long Videos</h2><h3 id="chunking-settings-for-longer-videos">⚙️ Chunking Settings for Longer Videos</h3>Adjust ffn_chunks based on your video length (at 24fps):<ul><li>10 seconds (240 frames): ffn_chunks=1-2</li><li>15 seconds (360 frames): ffn_chunks=2-4</li><li>20 seconds (480 frames): ffn_chunks=4-6</li><li>25 seconds (600 frames): ffn_chunks=8-10</li><li>33 seconds (800 frames): ffn_chunks=12-16</li></ul>If you encounter OOM (Out of Memory) errors:<ul><li>Increase ffn_chunks value</li><li>Reduce resolution slightly</li></ul>

LTX2 All-in-One ComfyUI Workflow

<ul><li>Fast Preview &gt; Upscale workflow optimization</li><li>Latent Normalization Node (This will improve audio/video quality by fixing overbaking and audio clipping issues.)</li><li>ComfyUI_LTX-2_VRAM_Memory_Management integration</li></ul>

Files with SHA256: 4d3206c73f7515ee4ada5e46b56cea3a0695d6ecf729718658eb02b578303da4