Hugging Face's logo
pgigi's picture

pgigi

pggigi
1

AI & ML interests

None yet

Recent Activity

repliedto dayona's post about 14 hours ago
Wan 2.2 I2V 14B Lightning Generate ultra-realistic, cinematic motion videos in 4–8 steps with 70%+ ZeroGPU quota savings ⚡ 4-Step Lightning Inference 🚀 FP8 Quantized & AoT C++ Compiled 🎬 Prompt Relay Multi-Event 👤 Lock Face Identity (CPU 0 Quota) 📥 Custom Civitai LoRA Downloader 🎬 RIFE 32 FPS (2s GPU ➔ 4s Video) ⚡ Wan 2.2 4-Step Lightning & AoT C++ Compilation Base Wan 2.2 14B model pre-compiled with NVIDIA C++ AOT kernels (~2.08s/step), rendering videos in 4 steps (10x faster than 40 steps). 🎬 Prompt Relay (Multi-Event Timeline Control) Exclusive feature routing multi-scene storylines ([0s-2s] Scene A ➔ [2s-4s] Scene B) in a single render with 0 extra GPU quota. 👤 Lock Face Identity (CPU Face Swap) Preserves and re-applies exact face identity from input image across frames using InsightFace on CPU (0 GPU quota used). 📥 Custom Civitai LoRA Downloader & CPU Pre-Download Paste any Civitai direct URL to download LoRAs to cache. Features a CPU Pre-Download button to prevent GPU quota waste on network drops. 🎬 RIFE 32 FPS Fluidity & Duration Extension Renders 2s on GPU (~8.3s quota), then CPU interpolates frames into a ultra-smooth 4s video with 0 extra GPU quota. 🎯 Smart ZeroGPU Quota Auto-Calibration Auto-calibrated GPU quota reservation formula adjusting for CFG scale & 3D attention, preventing premature quota exhaustion. https://huggingface.co/spaces/dayona/wan2.2-smart-i2v
liked a Space 1 day ago
dayona/I2V-EXTENDED
View all activity

Organizations

None yet