Endless Wan 2.2 Workflow Updated: Image-to-Video & Unlimited Duration
embryo10 · reddit · 2026-08-13
The Endless Wan 2.2 I2V workflow has been updated to v3.5, offering a simple solution to create Wan 2.2 videos of unlimited duration using SVI 2.0 Pro.
- Core Features: The workflow includes a 5-second "Initial" block and 8 optional 5-second "Extend" blocks, capable of producing up to 45 seconds of video. More time can be added by copying extend blocks.
- Generation Control: Generation can start from an initial image or an existing video. The Initial Image block now features an "End Frame" option to guide the generation from beginning to end.
- Performance Optimization: Each block has its own independent prompt selector, length control, fixed noise seed, and LoRA section. Users can experiment with specific blocks without regenerating previous ones.
- Hardware Compatibility: Supports GGUF loaders for low VRAM systems and Safetensors loaders. It also supports deeply optimized, distilled LoRAs (like Wan-Lightning) to generate high-quality video in as few as 4 steps.
More from Multimodal
- Video Generation Models Struggle with Complex Physical Interactions Like Handcuffing — flowersslop · 2026-08-13
- Testing LTX 2.5 Video Generation: A Children's Story Workflow — pfeifits · 2026-08-13
- GPT-4o Demonstrates Impressive 'Pixel Forensics' Capabilities — teortaxesTex · 2026-08-13
- Qualcomm VP on Image Gen Bottlenecks: Separating Scene Planning from Rendering — TWIML AI Podcast · 2026-08-13
- Runway Integrates xAI's Grok Imagine Image 2.0 Model — runwayml · 2026-08-13
- Video Diffusion Models Enable Single-Image Refocusing — CSProfKGD · 2026-08-13