invoke-ai/InvokeAI · error · ValueError

'latents' or 'noise' must be provided!

Error message

'latents' or 'noise' must be provided!

What it means

prepare_noise_and_latents needs a starting latent tensor: either an explicit `latents` field (image-to-image / img2img continuation) or a `noise` tensor (txt2img, from which zero latents are derived). If both are absent it cannot construct the initial sample and raises this error.

Source

Thrown at invokeai/app/invocations/denoise_latents.py:845

        Expected workflows:
        - Text-to-Image Denoising: `noise` is provided, `latents` is not. `latents` is initialized to zeros.
        - Image-to-Image Denoising: `noise` and `latents` are both provided.
        - Text-to-Image SDXL Refiner Denoising: `latents` is provided, `noise` is not.
        - Image-to-Image SDXL Refiner Denoising: `latents` is provided, `noise` is not.

        NOTE(ryand): I wrote this docstring, but I am not the original author of this code. There may be other workflows
        I haven't considered.
        """
        noise = None
        if noise_field is not None:
            noise = context.tensors.load(noise_field.latents_name)

        if latents_field is not None:
            latents = context.tensors.load(latents_field.latents_name)
        elif noise is not None:
            latents = torch.zeros_like(noise)
        else:
            raise ValueError("'latents' or 'noise' must be provided!")

        if noise is not None and noise.shape[1:] != latents.shape[1:]:
            raise ValueError(f"Incompatible 'noise' and 'latents' shapes: {latents.shape=} {noise.shape=}")

        # The seed comes from (in order of priority): the noise field, the latents field, or 0.
        seed = 0
        if noise_field is not None and noise_field.seed is not None:
            seed = noise_field.seed
        elif latents_field is not None and latents_field.seed is not None:
            seed = latents_field.seed
        else:
            seed = 0

        return seed, noise, latents

    def invoke(self, context: InvocationContext) -> LatentsOutput:
        if os.environ.get("USE_MODULAR_DENOISE", False):
            return self._new_invoke(context)

View on GitHub (pinned to 0b6a024f2f)

Solutions

1. Connect a Noise invocation (with seed) to the DenoiseLatents `noise` input for txt2img
2. Or supply `latents` (e.g. from a VAE-encode or previous DenoiseLatents output) for img2img
3. Ensure both fields aren't accidentally set to null in a JSON-built graph

Example fix

// before
DenoiseLatents(...)  # noise and latents both None
// after
noise = NoiseInvocation(seed=42)
denoise = DenoiseLatents(noise=noise, ...)

if latents_field is None and noise is None:
    raise ValueError("Provide either a noise field (txt2img) or a latents field (img2img) before invoking")

def has_init_sample(latents_field, noise) -> bool:
    return latents_field is not None or noise is not None

try:
    out = invocation.invoke(context)
except ValueError as e:
    if "'latents' or 'noise' must be provided" in str(e):
        invocation.noise = NoiseInvocation(seed=0)
        out = invocation.invoke(context)
    else:
        raise

When it happens

Trigger: Invoking DenoiseLatents with both `latents` and `noise` inputs unconnected/None — e.g. a graph where neither a NoiseInvocation output nor latents from VAE/previous denoise are wired in.

Common situations: Incomplete graphs in the canvas/editor where the noise node was deleted; API calls omitting both fields; copy-pasted partial workflows missing the noise branch.

Related errors

• Incompatible 'noise' and 'latents' shapes: ${latents.shape=}
• Unexpected T2I-Adapter base model type: '${t2i_adapter_model
• Negative conditioning is required when guidance_scale > 1.0
• denoising_start must be 0 when no initial latents are provid
• Latents to blend must be the same size.

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/ad7a35f66a25ab84. Report an issue: GitHub.