Instructions to use alibaba-pai/CogVideoX-Fun-V1.5-5b-InP with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- VideoX Fun
How to use alibaba-pai/CogVideoX-Fun-V1.5-5b-InP with VideoX Fun:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
CogVideoX-Fun V1.5 community docs proposal: asynchronous local video API on Mac
Proposing a community integration note for the V1.5 5B InP checkpoint served through Rapid-MLX's asynchronous video endpoint on Apple Silicon.
The useful client flow is capability discovery, creating a /v1/videos job, polling its status and fetching the completed content. The linked implementation tests cover a persistent worker, output encoding, cancellation cleanup and rejection of unsupported request shapes.
The “InP” checkpoint name should not imply that every upstream inpainting/control workflow is exposed by this API. A proposed example should specify its accepted conditioning, frame count and output size, then retain one successful video artifact with exact model/runtime pins. Q4, Q8 and BF16 are separate qualification targets. This is a request for a bounded community documentation entry, not a claim of ComfyUI feature parity.
Concrete catalog entry: cogvideox-fun-5b-q4 → dgrauet/CogVideoX-Fun-V1.5-5b-InP-mlx-q4@701df716bb7ae620ddccd109a1fbe8bb0a2a3538. This is a conversion/distribution artifact; it is not presented as an official upstream release.
Source references (fixed Rapid commit): catalog, video client guide, CogVideoX contracts.
Verification status: this proposal is based on reading the catalog, implementation and the linked documentation/tests. No fresh model inference or test-suite run was performed for this submission; source tests are not an end-to-end qualification. It requests a community documentation scope, not an official endorsement or a performance listing.
Disclosure: I maintain Rapid-MLX. Prepared with AI assistance under my direction. Anonymous usage-metadata telemetry is enabled by default; DO_NOT_TRACK=1 disables it. The original model and conversion terms continue to apply.