Instructions to use bkjha8/beenga-motion-5b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use bkjha8/beenga-motion-5b with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("bkjha8/beenga-motion-5b", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
Beenga Motion 5B
Text- and image-to-video base for the Beenga pipeline.
Part of Beenga Video OS โ a measured, licence-audited video pipeline built for Indian users.
| Role | Motion โ generates video from a prompt or a still |
| Parameters | 5B |
| Licence | Apache-2.0 |
| Attribution | see NOTICE |
Provenance
These weights are an unmodified redistribution of an upstream Apache-2.0 model.
Nothing has been trained, fine-tuned, quantised or converted. The upstream project,
the exact revision, and the original authors are recorded in the NOTICE file in
this repository, and the full licence text is in LICENSE.
Beenga hosts its own copy so that no part of its pipeline depends on a repository outside its control. Apache-2.0 grants a perpetual, irrevocable right to use and redistribute the weights as released โ it does not guarantee the upstream repository stays reachable, stays public, or keeps the same licence on future versions.
The Beenga name here identifies whose copy this is and what role it plays in the
pipeline, not authorship of the model. Credit for the weights belongs to the
upstream authors named in NOTICE.
Related
- Pipeline, benchmark and measurements: https://github.com/Beenga/beenga-video-os
- Inference and training code, also vendored: Beenga/Wan2.2, Beenga/DiffSynth-Studio
- Downloads last month
- 8