FlowMDM BABEL for Motius
This repository packages the official FlowMDM BABEL checkpoint, BABEL-135 normalization statistics, and inference metadata for the Motius pipeline.
from motius.pipelines.flowmdm import FlowMDMPipeline
pipe = FlowMDMPipeline.from_pretrained(
"ZeyuLing/motius-flowmdm-babel",
bundle_kwargs={"device": "cuda"},
device="cuda",
)
motion = pipe.infer_sequential_t2m(
[["walk forward", "turn left", "sit down"]],
[[60, 45, 75]],
)[0]
The output is an unnormalized (T, 135) BABEL motion. See the
Motius FlowMDM model card
for representation, conversion, and evaluation details.
FlowMDM source and weights are redistributed under the license included in
LICENSE. Original project: https://github.com/BarqueroGerman/FlowMDM
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support