Unverified50% confidenceFactExact time
一段16帧、分辨率256×256、patch大小16×16的视频送入Transformer会产生4096个token
1
Sources
50%
Confidence
Long-term
Relevance
8/29/2026
First Seen
Sources
Related Entities
Related Claims
Unverified512×512图像切分为32×32网格，共1024个token；1024分辨率则增至4096个token74% similarUnverifiedA raw video of 512×512 resolution and 24 frames has a data volume of approximately 18 MB, which is reduced to approximately 1/96 of the original after VAE compression into the 64×64×4 latent variable space.71% similarUnverifiedUnder certain configurations, graphics cards with 16GB of VRAM can also complete video generation with the Wan model at lower resolutions.70% similarUnverifiedModels such as Llama-3.1-8B are built on a stacked Transformer decoder architecture, typically containing 32 decoder layers, with approximately 250 million parameters per layer and approximately 500MB of VRAM per layer at FP16 precision.70% similarUnverified标准720p视频为1280×720（约92万像素），1080p为1920×1080（约207万像素）69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/816339API
curl https://kongchang.com/api/v1/knowledge/claims/816339MCP
get_claim(id=816339)