WorldFoundry
English中文

DynamiCrafter

Method overview: Flowchart of the proposed DynamiCrafter . During training, we randomly select a video frame as the image condition of the denoising process through the proposed dual-stream image injection mechanism to inherit visual.

Tencent / CUHK open-domain I2V. Two wired variants: 512 and 1024. Official long-config parity pending. Redistribution gated until license review.

Edit this pageGitHubMarkdown

DynamiCrafter animates an open-domain still: picture (plus optional prompt) in, a plausible motion clip out. It is a U-Net-era latent video diffusion with dual-stream image conditioning (CLIP image features + VAE latents). WorldFoundry wraps two resolution profiles as in-tree runtimes. Default pipeline: DynamiCrafter512I2VPipeline. Variants: dynamicrafter-512-i2v, dynamicrafter-1024-i2v.

Method overview: Flowchart of the proposed DynamiCrafter . During training, we randomly select a video frame as the image condition of the denoising process through the proposed dual-stream image injection mechanism to inherit visual.
Method overview: Flowchart of the proposed DynamiCrafter . During training, we randomly select a video frame as the image condition of the denoising process through the proposed dual-stream image injection mechanism to inherit visual.

Paper: arXiv 2310.12190. Project: doubiiu.github.io/projects/DynamiCrafter. Publisher: Tencent AI Lab / CUHK. Catalog license unset; the manifest gates redistribution until upstream license review.

Both variants are wired; official long-config parity is pending. Start at 512 for speed, 1024 for finals. It invents motion — it does not follow camera paths or actions.

Build your run

Pipeline
DynamiCrafter512I2VPipelineworldfoundry.pipelines.dynamicrafter.pipeline_dynamicrafter_512_i2v
TaskInteractive Video interactive-video
Environmentworldfoundry-unified-cu128
DeviceCUDA 12.8
Runnerworldfoundry.pipeline
Loadingfrom_pretrained
Invocationnative_or_standard_invocation
worldfoundry-eval run \
  dynamicrafter-512-i2v \
  --pipeline.task-profile interactive-video \
  --json

Variants

  • dynamicrafter-512-i2v
    StatusRunner parity pending

    image-to-video

    Runtime profiledynamicrafter-512-i2v

    Pipeline bindingdynamicrafter-512-i2v

  • dynamicrafter-1024-i2v
    StatusRunner parity pending

    image-to-video

    Runtime profiledynamicrafter-1024-i2v

    Pipeline bindingdynamicrafter-1024-i2v

What it is for

Animate a still with DynamiCrafter I2V.

Launch worldfoundry-eval run dynamicrafter-512-i2v.

Score on AIGCBench — image-to-video benchmark aligned with DynamiCrafter's task.

Runtime contract

Required: prompt and image; optional: video; actions: seed; artifacts: dynamicrafter-512-i2v.mp4.

VariantSizeFramesFrame strideArtifact
dynamicrafter-512-i2v320×5125024dynamicrafter-512-i2v.mp4
dynamicrafter-1024-i2v576×10245010dynamicrafter-1024-i2v.mp4

Run in WorldFoundry

Status: Runner parity pending. Unified environment worldfoundry-unified-cu128 (Python 3.11, CUDA 12.8). Stage Doubiiu/DynamiCrafter, Doubiiu/DynamiCrafter_512, and Doubiiu/DynamiCrafter_1024 locally — missing files fail at load, nothing is downloaded. Run worldfoundry-eval run dynamicrafter-512-i2v. Catalog VRAM is unset.