Unverified50% confidenceBenchmarkExact time
Attempts to use the video generation model Wan 2.2 TI2V to switch to a target character's perspective via instructions failed completely, though the developer has not yet tested larger-scale video models.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/14/2026
First Seen
Valid until: 12/13/2026
Sources
Related Entities
Related Claims
UnverifiedThe WAN 2.3 video generation model cannot recognize camera movement instructions, cannot understand action descriptions, and has poor prompt comprehension capabilities.74% similarUnverifiedWan 2.2 TI2V (Text-Image to Video) is a type of multimodal generative model that takes an image plus text as input and generates continuous video frames.69% similarUnverifiedLong Take mode is experimental in nature, may fail mid-generation with no retry mechanism, and the developer recommends pairing it with LTX Video 2.3 or Wan Video for best results.67% similarUnverifiedByteDance open-sourced the Bernini video editing model trained on Wan 2.2.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/919986API
curl https://kongchang.com/api/v1/knowledge/claims/919986MCP
get_claim(id=919986)