PKU-YuanGroup/ConsisID
ConsisID is a diffusion-based text-to-video generation system that preserves facial identity consistency across generated video sequences using frequency decomposition.
Not currently ranked — collecting fresh signals.
star history
ConsisID generates identity-preserving videos from text prompts by decomposing features across frequency domains to maintain facial consistency throughout generated videos. The system leverages diffusion models as its core generative architecture to produce coherent video sequences with faithful identity representation. This approach addresses a key challenge in video generation—maintaining subject identity across temporal frames—through technical innovations in how visual features are decomposed and processed.
Frequently asked
- What is PKU-YuanGroup/ConsisID?
- ConsisID is a diffusion-based text-to-video generation system that preserves facial identity consistency across generated video sequences using frequency decomposition.
- Is ConsisID open source?
- Yes — PKU-YuanGroup/ConsisID is open source, released under the Apache-2.0 license.
- What language is ConsisID written in?
- PKU-YuanGroup/ConsisID is primarily written in Python.
- How popular is ConsisID?
- PKU-YuanGroup/ConsisID has 848 stars on GitHub.
- Where can I find ConsisID?
- PKU-YuanGroup/ConsisID is on GitHub at https://github.com/PKU-YuanGroup/ConsisID.