The launch video does not render on the model card today. This PR fixes that.
## Why the current approach does not work
The `widget:` block added in #2 is the **Inference Widget** spec, not a general media embed. Its `output.url` only populates the example-output slot *inside* that widget. Per the Hub docs (`models-widgets.md`, "Widget Availability"), a widget requires the model's task to be supported by at least one Inference Provider. `pipeline_tag: robotics` has no widget and no provider support, so there is no widget to host the output. The Hub parses the `widgetData` (it is present verbatim in the rendered page payload) and then discards it. Silently, which is why it looked correct.
Confirmed this is tag-gating, not a problem with the asset:
- `calcuis/wan-gguf` (`pipeline_tag: text-to-video`), structurally identical widget YAML: renders a Gallery and real `<video>` elements.
- `PerceptronAI/Isaac-0.5` (`pipeline_tag: robotics`), same YAML shape: zero `<video>` elements in the…
- Add Isaac 0.5 launch video to the robotics Video Preview widget (a1be119f6c721de35482639e716ad9f44a9d26cc)
Co-authored-by: Eric Pence <ericpence@users.noreply.huggingface.co>
