Add Echo modular pipeline

[Gelercatty]

What does this PR do?

Adds official Diffusers Modular Pipeline support for Echo, a long-video model that generates synchronized video and audio with cross-shot memory.

This supersedes #13910 with a modular implementation that incorporates the earlier review feedback:

• adds the standalone diffusers.modular_pipelines.echo package and the public EchoModularPipeline / EchoBlocks APIs
• reuses existing lower-level Diffusers components without exposing LTX in Echo's public API
• keeps checkpoint conversion in scripts/ instead of adding an original-checkpoint runtime pipeline
• supports an optional first frame and up to seven paired image/audio memory slots
• supports variable-length memory audio and crops audio longer than 9.62 seconds to its highest-response window
• implements Echo's positive-only conditioning and stochastic 8-step DMD denoising workflow
• adds mixed-precision-safe audio decoding, documentation, and tests using a public tiny fixture

Official resources:

• Source implementation: https://github.com/jd-opensource/JoyAI-Echo/tree/echo-long-diffusers
• Model weights: https://huggingface.co/Echo-Team/JoyAI-Echo
• Tracking issue: [New Pipeline/Model] Add JoyAI-Echo multi-shot audio-video generation pipeline #13909

Closes #13909
Supersedes #13910

Testing

• make style
• make fix-copies
• make quality
• PYTHONPATH=src python -m pytest -q tests/modular_pipelines/echo/test_modular_pipeline_echo.py
  • 25 passed, 12 skipped

AI-assisted self-review

Codex was used to help review, refactor, test, and prepare this contribution. The final diff was reviewed using the repository's model-integration and self-review guidance.

The final self-review found no blocking correctness issues. Echo has its own public package and naming throughout; existing lower-level Diffusers component types are reused only as implementation dependencies. Memory-slot count, short-audio handling, variable audio lengths, response-aware 9.62-second cropping, deterministic DMD re-noising, mixed-precision decoding, loading, save/load, workflow, and memory behavior are covered by the test suite.

Validation scope: the final pass used the public Echo-Team/tiny-echo-modular-pipe fixture on CPU. A full production-checkpoint conversion and end-to-end GPU inference run was not repeated in this final pass.

Before submitting

• Did you use an AI agent (Claude Code, Codex, Cursor, etc.) to help with this PR?
  • Did you read the Coding with AI agents guide?
  • Did you run the self-review skill on the diff?
  • Did you share the final self-review notes in the PR description or a comment?
• Did you read the contributor guideline?
• Did you read our philosophy doc?
• Was this discussed/approved via a GitHub issue or the forum? See [New Pipeline/Model] Add JoyAI-Echo multi-shot audio-video generation pipeline #13909 and the maintainer discussion on [feat] Add JoyAI-Echo multi-shot audio-video generation pipeline #13910.
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?
• Are you the author (or part of the team) of the model/pipeline?

Who can review?

@asomoza @yiyixuxu

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[yiyixuxu]

thanks, I left some feedbacks

[yiyixuxu]

can you change file names to remove the echo_ prefix?

[yiyixuxu]

Suggested change

	components,
	tokenizer,
	text_encoder,

for module-level functions, let's take explicit argument instead of componennts

[yiyixuxu]

Suggested change

	OutputParam(
	"batch_size",
	type_hint=int,
	description="Number of prompts before per-prompt expansion.",
	),
	OutputParam("dtype", type_hint=torch.dtype, description="Prompt embedding dtype."),

so text_encoder step can be used on its own to compute prompt_embeds -> we want to minimize outputs, batch_size and dtype can easily derived later

[yiyixuxu]

let's move all the assembled blocks into modular_blocks_echo.py

[yiyixuxu]

can we in-line this function?

[yiyixuxu]

Suggested change

	components,
	transformer,

let's make methods/functions argument more explicit

[yiyixuxu]

Suggested change

	components,
	transformer,

[yiyixuxu]

can we keep only the vae encoding in this step, and move the part that requires transformer into a block in before_denoise.py? so that the encoder step do not have to declare transformer as a component

[yiyixuxu]

I think it's not needed, the code did not use it - also, vaes should only be required in enocder/decoder steps

Suggested change

ComponentSpec("audio_vae", AutoencoderKLLTX2Audio),

[yiyixuxu]

i think we can have a
text_encoder, "vae_encoder","denoise","decode"

see some write up on assembling the blocksets https://github.com/huggingface/diffusers/blob/main/.ai/references/modular.md#key-pattern-flat-blocksets