WorldFoundry
English中文

HunyuanVideo 1.5

Method overview: HunyuanVideo 1.5’s Diffusion Transformer for latent video generation.

Tencent HunyuanVideo 1.5 T2V/I2V on shared native infra; I2V needs gated FLUX.1-Redux-dev.

Edit this pageGitHubMarkdown

HunyuanVideo 1.5 is the newer revision of Tencent's open video family, again recorded with independent T2V and I2V recipes on one shared native infrastructure. The notable difference from the original: the I2V conditioner declares FLUX.1-Redux-dev as an independent gated vision checkpoint instead of inferring a sibling path.

Method overview: HunyuanVideo 1.5’s Diffusion Transformer for latent video generation.
Method overview: HunyuanVideo 1.5’s Diffusion Transformer for latent video generation.

Build your run

Pipeline
HunyuanVideo15T2VPipelineworldfoundry.pipelines.hunyuan_video.pipeline_hunyuan_video
TaskInteractive Video interactive-video
Environmentworldfoundry-unified-cu128
DeviceCUDA 12.8
Runnerworldfoundry.pipeline
Loadingfrom_pretrained
Invocationnative_or_standard_invocation
worldfoundry-eval run \
  hunyuanvideo-1.5-t2v \
  --pipeline.task-profile interactive-video \
  --json

Variants

  • hunyuanvideo-1.5-t2v
    StatusRunner parity pending

    text-to-video

    Runtime profilehunyuanvideo-1.5-t2v

    Pipeline bindinghunyuanvideo-1.5-t2v

  • hunyuanvideo-1.5-i2v
    StatusRunner parity pending

    image-to-video

    Runtime profilehunyuanvideo-1.5-i2v

    Pipeline bindinghunyuanvideo-1.5-i2v

What it is for

Start on hunyuanvideo-1.5-t2v. Launch worldfoundry-eval run hunyuanvideo-1.5-t2v. Use hunyuanvideo-1.5-i2v only after accepting the gated FLUX.1-Redux-dev terms.

VBench is the listed T2V suite.

Runtime contract

Required prompt; actions is seed. Default artifact hunyuanvideo-1.5-t2v.mp4. Variants: hunyuanvideo-1.5-t2v, hunyuanvideo-1.5-i2v. I2V conditioner declares black-forest-labs/FLUX.1-Redux-dev as an independent gated vision checkpoint.

Running in WorldFoundry

Unified worldfoundry-unified-cu128 (Python 3.11, CUDA 12.8). Stage tencent/HunyuanVideo-1.5. Main weights carry other license metadata. Status: Runner parity pending for native checkpoint-backed CUDA. Accept FLUX.1-Redux-dev terms on Hugging Face before I2V.