invoke-ai/InvokeAI · error · ValueError

Wan latents-to-image requires batch size 1; got {latents.sha

Error message

Wan latents-to-image requires batch size 1; got {latents.shape[0]}.

What it means

The node decodes exactly one image, so the latents batch dimension must be 1. Batched latents (>1) are rejected explicitly rather than decoding only the first frame silently, which would hide data loss.

Source

Thrown at invokeai/app/invocations/wan_latents_to_image.py:58

    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__}.")

        spatial_scale = getattr(vae_info.model.config, "scale_factor_spatial", None) or 8
        estimated_working_memory = estimate_vae_working_memory_wan(
            operation="decode",

View on GitHub (pinned to 0b6a024f2f)

Solutions

Slice off a single batch item: latents[0:1] before decoding
Run the node once per batch element instead of batching
If downstream expects images, add a per-item loop or use a batch-capable node

Example fix

// before
latents.shape == (4, C, H, W) -> error
// after
latents = latents[:1]  # or latents[i:i+1] per item

Defensive patterns

Strategy: validation

Validate before calling

if latents.shape[0] != 1:
    latents = latents[:1]  # or loop over batch items

Type guard

def is_single_batch(t: torch.Tensor) -> bool:
    return t.shape[0] == 1

Try / catch

try:
    out = wan_latents_to_image.invoke(context)
except ValueError as e:
    if 'batch size 1' in str(e):
        out = [decode_one(latents[i:i+1]) for i in range(latents.shape[0])]
    else:
        raise

Prevention

Avoid batching with Wan image decode; iterate instead
Slice latents[:1] when only one output is needed
Keep per-item processing loops for batch workflows

When it happens

Trigger: Feeding latents with shape[0] > 1 (e.g., from a batched generation or a manual torch.stack of multiple latent tensors) into Wan Latents to Image.

Common situations: Batch workflows built for txt2img SD pipelines reused for Wan; users stacking multiple encoded images into one tensor.

Related errors

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/4eecd02dfa75f0b2. Report an issue: GitHub.