[New Pipeline/Model] Add JoyAI-Echo multi-shot audio-video generation pipeline

[sjq66]

Model/Pipeline/Scheduler description

We are the JoyAI Team (JD.com), proposing the integration of JoyAI-Echo into Diffusers.

JoyAI-Echo is a unified framework for long-form audio-visual generation, designed to support minute-level video creation with synchronized audio, strong temporal consistency, and real-time interaction.

Key innovations:

• Cross-modal audio-visual memory bank: preserves character appearance and voice timbre across long sequences (up to minutes)
• DMD-distilled few-step inference: ~7.5× faster than baseline while improving alignment and visual quality
• Joint audio-video generation: a single pipeline produces synchronized video and audio
• Multi-shot story generation: generates coherent sequences of shots from prompt lists

The architecture builds on LTX-2 and adds the JoyAI-Echo DMD denoising schedule plus a paired audio-video memory bank for cross-shot consistency.

Open source status

• The model implementation is available.
• The model weights are available (Only relevant if addition is not a scheduler).

Provide useful links for the implementation

• GitHub Repository: https://github.com/jd-opensource/JoyAI-Echo
• HuggingFace Weights: https://huggingface.co/jdopensource/JoyAI-Echo
• Diffusers implementation PR: (will link after PR is created)

Additional context

We (JoyAI Team) previously contributed JoyAI-Image-Edit to Diffusers (PR #13444, merged). This follows the same pattern — official team providing a complete, tested implementation.