invoke-ai/InvokeAI · error · ValueError

Wan latents-to-image expects a 4D or 5D latent tensor [B, C,

Error message

Wan latents-to-image expects a 4D or 5D latent tensor [B, C, (T), H, W]; got {tuple(latents.shape)}.

What it means

Wan Latents to Image decodes a single image and expects the latents tensor to be 4D [B,C,H,W] or 5D [B,C,T,H,W]. Any other rank (2D, 3D, 6D, etc.) cannot be interpreted, so the node raises with the actual shape. This catches feeding incompatible latents from other pipelines.

Source

Thrown at invokeai/app/invocations/wan_latents_to_image.py:54

    "wan_l2i",
    title="Latents to Image - Wan 2.2",
    tags=["latents", "image", "vae", "l2i", "wan"],
    category="latents",
    version="1.0.0",
    classification=Classification.Prototype,
)
class WanLatentsToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
    """Decodes Wan latents back to RGB."""

    latents: LatentsField = InputField(description=FieldDescriptions.latents, input=Input.Connection)
    vae: VAEField = InputField(description=FieldDescriptions.vae, input=Input.Connection)

    @torch.no_grad()
    def invoke(self, context: InvocationContext) -> ImageOutput:
        latents = context.tensors.load(self.latents.latents_name)

        if latents.ndim not in (4, 5):
            raise ValueError(
                f"Wan latents-to-image expects a 4D or 5D latent tensor [B, C, (T), H, W]; got {tuple(latents.shape)}."
            )
        if latents.shape[0] != 1:
            raise ValueError(f"Wan latents-to-image requires batch size 1; got {latents.shape[0]}.")

        # This node decodes exactly one image. Multi-frame video latents would otherwise
        # run the full (expensive) multi-frame VAE decode — under a working-memory
        # estimate that assumed one frame — and then die in an opaque einops rank error
        # at the final rearrange. Checked before the VAE is even loaded.
        if latents.ndim == 5 and latents.shape[2] != 1:
            raise ValueError(
                f"These latents hold {latents.shape[2]} frames of video; this node decodes a single "
                "image. Use 'Latents to Video - Wan 2.2' (wan_l2v) for video latents."
            )

        vae_info = context.models.load(self.vae.vae)
        if not isinstance(vae_info.model, AutoencoderKLWan):
            raise TypeError(f"Expected AutoencoderKLWan for Wan VAE, got {type(vae_info.model).__name__}.")

View on GitHub (pinned to 0b6a024f2f)

Solutions

  1. Supply latents from Wan Image to Latents or a Wan denoise node ([B,C,H,W] or [B,C,1,H,W])
  2. Check the upstream node type — replace non-Wan latents-to-image with the Wan one
  3. Inspect the tensor with tensor.shape; reshape/pad to rank 4 or 5 before decoding

Example fix

// before
latents.shape == (C, H, W)  // ndim=3 -> error
// after
latents = latents.unsqueeze(0)  // (1, C, H, W)
Defensive patterns

Strategy: type-guard

Validate before calling

if latents.ndim not in (4, 5):
    raise ValueError(f"expected 4D/5D latents, got ndim={latents.ndim} shape={tuple(latents.shape)}")

Type guard

def is_valid_wan_latents(t: torch.Tensor) -> bool:
    return t.ndim in (4, 5) and t.shape[0] == 1

Try / catch

try:
    out = wan_latents_to_image.invoke(context)
except ValueError as e:
    if '4D or 5D latent tensor' in str(e):
        latents = latents.unsqueeze(0)  # reshape as appropriate
    else:
        raise

Prevention

When it happens

Trigger: Loading latents produced by a non-Wan node or with an unexpected rank into the Wan latents-to-image node; passing preview/noise tensors or manually truncated tensors whose ndim is not 4 or 5.

Common situations: Wiring standard SD latents-to-image tensors into the Wan node; corrupt or hand-edited tensor files; intermediate debug tensors with squeezed batch/channel dims.

Related errors

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/87e0c15cdcb6365c. Report an issue: GitHub.