OPSD-V Checkpoints

This repository provides the checkpoints for four models: Self-Forcing, Self-Forcing + OPSD-V, LongLive, and LongLive + OPSD-V.

Download

hf download anonymousopsd-v/OPSD-V --local-dir .

Checkpoint files

Model Base generator LoRA adapter
Self-Forcing checkpoints/self_forcing_dmd_ema_as_generator.pt None
Self-Forcing + OPSD-V checkpoints/self_forcing_dmd_ema_as_generator.pt checkpoints/opsdv_self_forcing_lora.pt
LongLive checkpoints/longlive_base.pt checkpoints/longlive_lora.pt
LongLive + OPSD-V checkpoints/longlive_base.pt checkpoints/opsdv_longlive_lora.pt

The four models share two base generators, so these five files cover their model-specific checkpoints. Shared runtime components such as the VAE and text encoder are obtained separately as described in the code release. For LongLive + OPSD-V, load the OPSD-V adapter directly on the base generator; do not stack it with the original LongLive adapter.

Usage

Place the five files in the checkpoints/ directory of the anonymous code release. Use the corresponding configuration:

Model Configuration
Self-Forcing configs/inference_self_forcing_original.yaml
Self-Forcing + OPSD-V configs/inference_self_forcing.yaml
LongLive configs/inference_longlive_original.yaml
LongLive + OPSD-V configs/inference_longlive.yaml

Use the default checkpoint selection in these configurations. The Self-Forcing base checkpoint already stores its EMA-derived generator under the generator key; its filename does not require the --use_ema flag. The OPSD-V configurations load the online LoRA weights by default.

Visual results · Code and inference instructions

Upstream models

The base generators and original LongLive adapter come from the official Self-Forcing and LongLive releases. Please follow the licenses and usage terms of the corresponding upstream checkpoints.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support