Wayne-King/echo-memory-diffusers

Echo-Memory Diffusers pipeline

Community pipeline that overlays the Echo-Memory context_k1 row onto official Wan 2.1 1.3B Diffusers weights.

• Paper: arXiv:2606.09803
• Code: Echo-Team-Joy-Future-Academy-JD/Echo-Memory
• Original keys: Echo-Team/Echo-Memory context_k1/epoch-0.safetensors
• Converted transformer: this repo, context_k1-diffusers/diffusion_pytorch_model.safetensors

This is not the full multi-chunk camera-action / SSM research stack. It is the released DiT fine-tune remapped to Diffusers names (825 / 825 official Wan 1.3B transformer keys).

import torch
from diffusers import DiffusionPipeline
from diffusers.utils import export_to_video

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 on the fly
# or: pipe.load_converted_echo_memory_weights()  # already-remapped transformer in this repo
pipe.to("cuda")

frames = pipe(
    prompt="A golden retriever running across a sunny green field, cinematic camera follow.",
    negative_prompt="blurry, static, low quality, deformed",
    height=480,
    width=832,
    num_frames=33,
    num_inference_steps=30,
    guidance_scale=5.0,
).frames[0]
export_to_video(frames, "echo_memory_context_k1.mp4", fps=16)