1.0.11

What's New

• Improved trajectory accuracy by normalizing interaction coordinates against the resolved media dimensions, keeping touch positions consistent across FIT/FILL rendering and downlink resolution changes.
• Improved image and local video sizing: preserve the media aspect ratio, avoid enlarging small inputs, cap input resolution at the model's pixel limit, and select the model generation canvas independently.
• Reduced trajectory signaling latency by retaining only the latest pending position for each generation task. Starting, changing, or stopping generation clears obsolete pending trajectories without interrupting in-progress message fragments.
• Updated the Agora signaling budget to 58 packets and 29,000 bytes per second while preserving control-message ordering.
• Updated XLab's touch-animation mode to use an empty default prompt and retain the uploaded input image as its reference image.
• Moved public README assets into materials/ and excluded internal development documentation from the repository.

API Changes

• Removed the optional videoFormat parameter from all createLocalImageStream(...) overloads and from createLocalVideoStream(uri). The SDK now derives the input format from the media and model defaults. Remove this argument from existing calls and recompile applications when upgrading.
• Camera, external video, and network video stream creation APIs are unchanged.
• Supported production models remain X2_1_1_PREVIEW_CN and X2_1_1_PREVIEW_US; the temporary drag test model is not included in this release.