[Community pipeline] Echo-Memory overlay for Wan 2.1 1.3B

[WayneJin0918]

Model/Pipeline/Scheduler description

Request to add a GitHub community pipeline that overlays the released Echo-Memory context_k1 DiT fine-tune onto official Diffusers Wan 2.1 1.3B.

This is not a new official Wan checkpoint and not the full Echo-Memory multi-chunk / camera-action / SSM research stack. It remaps original DiffSynth/Wan transformer keys onto WanPipeline.transformer (825 / 825 official keys) so users can load:

from diffusers import DiffusionPipeline

pipe = DiffusionPipeline.from_pretrained(
    "Wan-AI/Wan2.1-T2V-1.3B-Diffusers",
    custom_pipeline="pipeline_echo_memory",
    torch_dtype=torch.bfloat16,
)
pipe.load_echo_memory_weights()

PR: #14471

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

• Paper: https://arxiv.org/abs/2606.09803
• Code: https://github.com/Echo-Team-Joy-Future-Academy-JD/Echo-Memory
• Original keys: https://huggingface.co/Echo-Team/Echo-Memory (context_k1/epoch-0.safetensors)
• Converted transformer + Hub custom pipeline: https://huggingface.co/Wayne-King/echo-memory-diffusers
• Authors: @WayneJin0918 / Echo Team

[github-actions]

Thanks for the request!

How new model support works in Diffusers

We're a small team, and our review queue shouldn't be what decides whether a model is usable in Diffusers. With Modular Diffusers, a pipeline can live as remote code in any Hub repo and load straight from there with from_pretrained.

🛠️ Want to bring this model to Diffusers?

Please start with a Hub repo — you don't need anything from us to do that, and people can use it immediately. From there we decide how to support it: we might work with the authors, upstream an existing community version, or just point people at the one on the Hub. The pipelines we integrate are usually the ones people are already running.

Tag @asomoza when you have something to share — we'll give feedback on the implementation, help get it in front of people, and add the ones we like to our hand-picked Modular Pipelines collection. Tell us where you hit friction along the way, too: confusing APIs, missing docs, bugs. That feedback is worth as much to us as the pipeline.

👋 Are you an author of the model? We'd love to hear from you — comment here and we'll help you pick the path that fits.

📚 Quickstart · Building custom blocks — template repo, and how to publish to the Hub · Modular Pipelines and Custom Blocks — examples to crib from

This is an automated message.

[WayneJin0918]

I'm one of the Echo-Memory authors (@WayneJin0918). Thanks for the pointer — a Hub repo is already live, so people can use this without waiting on a GitHub merge:

https://huggingface.co/Wayne-King/echo-memory-diffusers

import torch
from diffusers import DiffusionPipeline

pipe = DiffusionPipeline.from_pretrained(
    "Wan-AI/Wan2.1-T2V-1.3B-Diffusers",
    custom_pipeline="Wayne-King/echo-memory-diffusers",
    torch_dtype=torch.bfloat16,
    trust_remote_code=True,
)
pipe.load_echo_memory_weights()  # remaps Echo-Team/Echo-Memory `context_k1`
# or: pipe.load_converted_echo_memory_weights()

What this is:

• official Wan 2.1 1.3B Diffusers pipeline + one released DiT row (context_k1)
• 825 / 825 transformer keys remapped from original DiffSynth / Wan names
• not the full multi-chunk / camera-action / SSM research stack

@asomoza happy to take whichever path you prefer:

1. Keep it as a Hub custom_pipeline (already usable)
2. Port to Modular Diffusers custom blocks if you want it in the collection
3. Keep or close the examples/community PR Add Echo-Memory community pipeline for Wan 2.1 1.3B #14471

Friction so far: Echo-Memory checkpoints use original Wan / DiffSynth keys, so a remap (or a converted file) is required before WanPipeline.transformer.load_state_dict. scripts/convert_wan_to_diffusers.py is not importable from a community file, so we duplicated the T2V rename dict.