psd-layer-vae g3

Use it to encode and decode layered images when sharper perceptual output during generation is preferred over maximum pixel fidelity.

Availability

Repository is public; access to files requires agreeing to conditions and sharing Hugging Face information.

Pricing model

Not established

Deployment

Inference code and weights are provided; example uses CUDA.

Privacy

Not established

Licence

Apache License 2.0 for weights and inference code.

• The g3 variant has a latent representation different from h3, and their encoders and decoders cannot be mixed. ^[1]
• The g3 variant is intended for sharper decoding during generation. ^[1]
• The source reports g3 rFID as half that of h3, with pixel fidelity about 0.3 dB lower. ^[1]
• The repository is publicly accessible, but files require agreeing to conditions and sharing Hugging Face information. ^[1]

  This repository is publicly accessible, but you have to accept the conditions to access its files and content .

• The source says g3's realistic-looking details do not necessarily match the original and recommends h3 for precise layer-by-layer reconstruction. ^[1]