Comfy-Org/ComfyUI · error · ValueError

SeedVR2TemporalChunk: expected {SEEDVR2_LATENT_CHANNELS} lat

Error message

SeedVR2TemporalChunk: expected {SEEDVR2_LATENT_CHANNELS} latent channels; got shape {tuple(samples.shape)}.

What it means

After the rank check, SeedVR2TemporalChunk verifies the latent has 16 channels (SEEDVR2_LATENT_CHANNELS) in dim 1, matching the SeedVR2 VAE. A 5-D latent from a different model family (or wrong VAE) fails here with the full shape.

Source

Thrown at comfy_extras/nodes_seedvr.py:463

            ],
            outputs=[
                io.Latent.Output(display_name="latents", is_output_list=True,
                                 tooltip="The temporal chunks in sequence order."),
                io.Int.Output(display_name="temporal_overlap",
                              tooltip="The effective latent-frame overlap between adjacent chunks, for Merge SeedVR2 Latents."),
            ],
        )

    @classmethod
    def execute(cls, latent, temporal_overlap, chunking_mode) -> io.NodeOutput:
        samples = latent["samples"]
        if samples.ndim != 5:
            raise ValueError(
                f"SeedVR2TemporalChunk: expected a 5-D video latent (B, C, T, H, W); "
                f"got shape {tuple(samples.shape)}."
            )
        if samples.shape[1] != SEEDVR2_LATENT_CHANNELS:
            raise ValueError(
                f"SeedVR2TemporalChunk: expected {SEEDVR2_LATENT_CHANNELS} latent channels; "
                f"got shape {tuple(samples.shape)}."
            )
        if temporal_overlap < 0:
            raise ValueError(
                f"SeedVR2TemporalChunk: temporal_overlap must be >= 0; got {temporal_overlap}."
            )
        mode = chunking_mode["chunking_mode"]
        if mode not in ("auto", "manual"):
            raise ValueError(
                f"SeedVR2TemporalChunk: chunking_mode must be 'auto' or 'manual'; "
                f"got {mode!r}."
            )
        t_latent = samples.shape[2]
        t_pixel = 4 * (t_latent - 1) + 1

        if mode == "auto":
            free_gb = comfy.model_management.get_free_memory(

View on GitHub (pinned to 1c6d8d45b3)

Solutions

Use the SeedVR2 video VAE for encoding so the latent is (B, 16, T, H, W).
Check the chain: the latent fed to Split must come from SeedVR2 encode/sample nodes.
Print samples.shape[1]; anything other than 16 means the wrong VAE/model produced it.

Defensive patterns

Strategy: validation

Validate before calling

samples = latent['samples']
if samples.ndim != 5 or samples.shape[1] != 16:
    raise ValueError(f'need SeedVR2 latent (B,16,T,H,W), got {tuple(samples.shape)}')

Type guard

def is_seedvr_chunk_input(s) -> bool:
    return s.ndim == 5 and s.shape[1] == 16

Prevention

Use one model family's VAE consistently per workflow.
Verify channel count (16) before chunking.
Don't adapt other architectures' latents with unsqueeze hacks.

When it happens

Trigger: Passing a 5-D latent whose channel dim != 16 — e.g. a 48-channel video VAE latent from another architecture, or a 4-channel image latent with extra dims unsqueezed.

Common situations: Mixing video model pipelines (Wan/Hunyuan latents into the SeedVR2 chunker); loading a non-SeedVR2 VAE for the encode step.

Related errors

AI-assisted analysis of Comfy-Org/ComfyUI@1c6d8d45b3 (2026-08-14). Data as JSON: /api/errors/b568b61aa6b3993e. Report an issue: GitHub.