Skip to content
Video

UniVOE

редактирование видео с учётом взаимодействий объектов

2 min read

UniVOE — фреймворк для универсального редактирования объектов в видео, построенный на Wan2.1-VACE-14B. Удаление и вставка объекта рассматриваются как взаимные операции: модель меняет не только сам объект, но и его следы в сцене — тени, отражения, отпечатки и освещение. Проект от Peking University, Tencent Hunyuan и The Chinese University of Hong Kong; полная система насчитывает 23B параметров, но в репозитории лежат 3B control branch и 153M DiT LoRA, а 14B backbone нужно скачивать отдельно.

Ключевые особенности

  • Удаление и инъекция объектов: remove-12step, remove-50step и inject-50step.
  • Effect-aware remover восстанавливает всю сцену без объекта, а не только область маски.
  • 12-step дистилляция: быстрый remover с rank-32 LoRA; 50-step примерно в 4 раза медленнее.
  • Инъекция без маски: объект может появиться в любом месте видео.
  • GRAR post-training обновляет injector онлайн по групповым относительным наградам.
  • До 237 кадров при 832×480; инструкции на английском.

Пример

huggingface-cli download Ysz2022/UniVOE --local-dir checkpoints/UniVOE
bash run_erase.sh
STEPS=50 bash run_erase.sh
bash run_inject.sh
ПодробнееПолная система: 23B параметров — 14B frozen Transformer, 3B control branch, text encoder/VAE. Файлы control branch весят по 6.1 GB в bf16, DiT LoRA — 307 MB.
Для removal дополнительно нужны GroundingDINO, SAM2 и bert-base-uncased; базовую Wan2.1-VACE-14B скачивают отдельно.
Ограничения: артефакты в crowded scenes и при сильном движении камеры; качество зависит от инструкции и reference image. Лицензия весов указана как LICENSE-PLACEHOLDER.
Discuss Telegram 1 429

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # Ysz2022/UniVOE
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download Ysz2022/UniVOE --local-dir ~/models/UniVOE
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to UniVOE
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • SageAttention 3v2.2.0from October 28, 2025

    even faster on RTX 50 / Blackwell

    FP4 attention for Blackwell. Faster than SageAttention 2, but the authors suggest v2 when accuracy matters.

    How to enable: Build from the sageattention3_blackwell folder of the repo; in ComfyUI via KJNodes.

    Earlier generations (2)
    • SageAttention 2++ — Quantised (INT8/FP8) attention with no visible quality loss. The main speed-up for image and video generation.
    • SageAttention 1 — First version: INT8 for Q·K. Still works, but 2++ is faster and more accurate.
  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (3)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
    • xFormers — The older way to speed up attention, works almost everywhere. Sage or Flash are faster where available.
  • cache-dit (DBCache)v1.5.6from November 20, 2026new release

    1.5–3× at nearly the same quality

    Reuses block outputs between diffusion steps. Supports most DiT models.

    How to enable: pip install cache-dit; one line cache_dit.enable_cache(pipe) for diffusers.

    Earlier generations (2)
    • MagCache — Skips steps by magnitude of change. More accurate than TeaCache, no calibration.
    • TeaCache — The first popular diffusion step cache. Works, newer methods are more accurate.
  • LightX2V step distill0.6.1from November 9, 2026new release

    5–10× fewer steps

    Distilled LoRAs and an engine: video in 4–8 steps instead of 30–50.

    How to enable: Load their step-distill LoRA, CFG 1.0, 4–8 steps.

  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of UniVOE (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

Ysz2022/UniVOE
Author
Ysz2022
Category
Video
Released
Posted

On NeyroLearn

Hugging Face

4291

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments

Video

Wan 2.2

Wan 2.2