invoke-ai/InvokeAI · error · ValueError

Wan latents-to-video requires non-empty temporal and spatial

Error message

Wan latents-to-video requires non-empty temporal and spatial dimensions.

What it means

A 5D latent tensor whose temporal (T) or spatial (H/W) dimensions contain a zero cannot be decoded into a video, so invoke() rejects it early with this ValueError. This guards the VAE against empty video tensors that would produce zero or undefined frames.

Source

Thrown at invokeai/app/invocations/wan_latents_to_video.py:103

        ge=1,
        le=120,
        description="Frames-per-second for the encoded MP4. Wan 2.2 was trained at 16 FPS.",
    )

    @torch.no_grad()
    def invoke(self, context: InvocationContext) -> VideoOutput:
        latents = context.tensors.load(self.latents.latents_name)
        _validate_video_latent_batch(latents)
        if latents.ndim == 4:
            # Promote 4D (single-frame) to 5D so this node can also serve as a
            # one-frame "video" encode if someone wires it that way.
            latents = latents.unsqueeze(2)
        if latents.ndim != 5:
            raise ValueError(
                f"Wan latents-to-video expects a 5D latent tensor [B, C, T, H, W]; got {tuple(latents.shape)}."
            )
        if any(size == 0 for size in latents.shape[2:]):
            raise ValueError("Wan latents-to-video requires non-empty temporal and spatial dimensions.")

        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__}.")

        if latents.shape[1] != vae_info.model.config.z_dim:
            raise ValueError(
                f"Latent channel mismatch: these latents have {latents.shape[1]} channels but the "
                f"selected VAE expects {vae_info.model.config.z_dim}. A14B models need the 16-channel Wan 2.1 VAE; "
                "TI2V-5B needs the 48-channel Wan 2.2 VAE."
            )

        _, _, t_lat, h_lat, w_lat = latents.shape
        spatial_scale = getattr(vae_info.model.config, "scale_factor_spatial", None) or 8
        temporal_scale = getattr(vae_info.model.config, "scale_factor_temporal", None) or 4
        t_pixel = (t_lat - 1) * temporal_scale + 1
        h_pixel, w_pixel = h_lat * spatial_scale, w_lat * spatial_scale
        optimize_memory = context.config.get().wan_memory_optimization

View on GitHub (pinned to 0b6a024f2f)

Solutions

  1. Fix the upstream computation/slicing so T, H, W are all >= 1.
  2. Validate latents.shape[2:] before invoking and abort the workflow branch early.
  3. Check that source image/video dimensions are non-zero before latent generation.

Example fix

// before
latents = denoise(prompt, num_frames=0)  # T == 0
video = wan_latents_to_video(latents=latents)  # ValueError
// after
assert latents.shape[2] > 0 and latents.shape[3] > 0 and latents.shape[4] > 0
video = wan_latents_to_video(latents=latents)
Defensive patterns

Strategy: validation

Validate before calling

if latents.ndim == 5 and any(s == 0 for s in latents.shape[2:]):
    raise ValueError(f"Wan video latents have an empty T/H/W dim: {tuple(latents.shape)}")

Type guard

def has_nonempty_video_dims(t) -> bool:
    return t.ndim == 5 and all(s > 0 for s in t.shape[2:])

Try / catch

try:
    video = node.invoke(context)
except ValueError as e:
    if "non-empty temporal and spatial" in str(e):
        raise WorkflowSkip("video branch produced empty latents")
    raise

Prevention

When it happens

Trigger: Passing latents where any of shape[2], shape[3], shape[4] is 0 — usually the result of an upstream slicing bug, an empty sequence after frame dropping, or an arithmetic error in latent size computation (e.g. (0 pixels) // downscale).

Common situations: Off-by-one slicing producing T=0; failed upstream resize leaving H or W at 0; conditional pipelines that emit placeholder empty tensors when a branch is skipped.

Related errors

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/ed38e4ffd0d3d4af. Report an issue: GitHub.