Skip to content
← Models
Open weightsAlibaba PAI/10 trustVerified

EasyAnimate

Alibaba's open DiT video pipeline for text-to-video, image-to-video, inpainting, and camera-controlled generation, up to V5.1 12B.

Visit official page →

About

EasyAnimate is Alibaba's Platform of AI (PAI) team entry in open video generation, presented in the May 2024 paper 'EasyAnimate: A High-Performance Long Video Generation Method based on Transformer Architecture'. It extends the diffusion transformer with a Hybrid Motion Module (temporal plus global attention) and a Slice VAE that compresses the temporal axis, originally targeting long videos of up to 144 frames. The line has matured through V5 to the current V5.1 12B checkpoints, which split into text-to-video, inpainting (image/video-to-video), control (Canny, depth, pose, trajectory), and camera-control variants, generating 1 to 49 frames at 8fps across 256 to 1024px resolutions.

It is one of the better-supported open video stacks: official checkpoints on Hugging Face under alibaba-pai, a diffusers pipeline, Alibaba Cloud PAI notebook tutorials, and even Huawei Ascend NPU ports. The repo is Apache-2.0 with 2.3k stars and was updated in October 2026. There are no published benchmark scores, but the breadth of modes (especially camera control and inpainting) makes it a practical choice for controllable open video work.

Specs

Context window

1 to 49 frames at 8fps (~6s); 256 to 1024px; T2V, I2V, V2V, inpainting, and control-to-video

Pricing

Free and open source (Apache-2.0); self-hosted or Alibaba Cloud PAI notebooks (cloud GPU costs apply)

Free quota

Full access, no quota, self-hosted

API access

No official API; weights on Hugging Face and ModelScope, diffusers pipeline, PAI notebooks

Weights

Open weights

Released

Not disclosed

Benchmarks

No published benchmark scores

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

controllable video generationcamera-move controlvideo inpaintingpose/depth-guided animation

Trust score — 8/10

/10 trustVerified

Scored 8/10 (metrics tier, metrics-v1). Evidence: Alibaba PAI paper arXiv 2405.18991 (May 2024); GitHub aigc-apps/EasyAnimate (2.3k stars, Apache-2.0 confirmed via GitHub API, updated Oct 2026); V5.1 12B checkpoints on HF (alibaba-pai) with diffusers support; Alibaba Cloud PAI docs. Strengths: unusually broad control modes, first-party cloud tutorials, permissive license. Docked for: (1) no published benchmark scores; (2) 12B checkpoints are heavy for consumer GPUs. License: Apache-2.0, confirmed via GitHub API.

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
8.0/10
Maintenance · recency, releases, issue load
8.0/10
Openness · license + docs
9.0/10
Manual review · sanity read + red-flag check
8.0/10
  • · Alibaba PAI; paper arXiv 2405.18991 (May 2024); 2.3k stars
  • · V5.1 12B checkpoints: T2V, InP, Control, Control-Camera; 1-49 frames at 8fps
  • · Diffusers pipeline, Alibaba Cloud PAI tutorials, Huawei Ascend ports
  • · No published benchmarks; 12B checkpoints heavy for consumer GPUs

Sources