Skip to content

Multi-Camera Path Control via Trajectory Conditioning Title: [Feature Request]: Trajectory-based Camera Path Conditioning for LTX-2 #219

@reeflose88-netizen

Description

@reeflose88-netizen

Describe the feature you'd like to see

I would love to see an extension to the core conditioning pipelines that allows users to supply structured camera trajectories (e.g., pan left/right, zoom in/out, orbital tracking) during inference. This could be implemented via extra conditioning inputs or an integrated spatial-temporal attention mechanism (similar to MotionCtrl or CameraCtrl implementations in other DiT architectures).

Is your feature request related to a problem?

Currently, controlling exact camera motion in text-to-video or image-to-video generation relies heavily on text prompting, which frequently results in unpredictable camera behaviors, unwanted "Ken Burns" drifting effects, or static frames.

Describe alternatives you've considered

Using external IP-Adapter or ControlNet hacks post-generation, which often breaks visual consistency or requires heavy reprocessing.

Additional context

Providing a clean python API wrapper within ltx-pipelines for coordinate-based or preset-based camera trajectories would significantly improve usability for professional cinematic workflows.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions