invoke-ai/InvokeAI · error · ValueError

denoising_start should be 0 when initial latents are not pro

Error message

denoising_start should be 0 when initial latents are not provided.

What it means

When no initial latents are supplied, FLUX denoising starts from pure noise, so denoising_start must be 0 — there is nothing to partially skip. A denoising_start greater than 1e-5 in a txt2img run is contradictory and raises this ValueError.

Source

Thrown at invokeai/app/invocations/flux_denoise.py:334

            if self.add_noise:
                assert noise is not None
                # Noise the orig_latents by the appropriate amount for the first
                # timestep in InvokeAI's clipped schedule.
                #
                # Known limitation: if the selected scheduler later replaces this
                # schedule with its own first effective timestep/sigma (for example
                # Heun internal expansion or LCM's scheduler-defined schedule), the
                # img2img preblend below may not match that scheduler's true first
                # step exactly. This is an existing pipeline limitation and affects
                # both internally generated noise and externally supplied noise.
                t_0 = timesteps[0]
                x = t_0 * noise + (1.0 - t_0) * init_latents
            else:
                x = init_latents
        else:
            # init_latents are not provided, so we are not doing image-to-image (i.e. we are starting from pure noise).
            if self.denoising_start > 1e-5:
                raise ValueError("denoising_start should be 0 when initial latents are not provided.")

            assert noise is not None
            x = noise

        # If len(timesteps) == 1, then short-circuit. We are just noising the input latents, but not taking any
        # denoising steps.
        if len(timesteps) <= 1:
            return x

        if is_schnell and self.control_lora:
            raise ValueError("Control LoRAs cannot be used with FLUX Schnell")

        # Prepare the extra image conditioning tensor (img_cond) for either FLUX structural control or FLUX Fill.
        img_cond: torch.Tensor | None = None
        is_flux_fill = transformer_config.variant is FluxVariantType.DevFill
        if is_flux_fill:
            img_cond = self._prep_flux_fill_img_cond(context, device=device, dtype=inference_dtype)
        else:

View on GitHub (pinned to 0b6a024f2f)

Solutions

1. Set denoising_start to 0 when no initial latents are provided
2. If you intend img2img, supply the initial latents input to the denoise node
3. Check upstream denoising_start/scheduler param nodes for leftover nonzero values

Example fix

// before: txt2img with img2img setting
FluxDenoiseInvocation(noise=noise, denoising_start=0.5)
// after
FluxDenoiseInvocation(noise=noise, denoising_start=0.0)

if initial_latents is None and denoising_start > 1e-5:
    raise ValueError('Set denoising_start=0 for txt2img, or provide initial latents for img2img')

try:
    result = flux_denoise.invoke(context)
except ValueError as e:
    if 'denoising_start should be 0' in str(e):
        flux_denoise.denoising_start = 0.0
        result = flux_denoise.invoke(context)
    else:
        raise

When it happens

Trigger: _run_diffusion takes the init_latents-is-None branch while self.denoising_start > 1e-5; running a text-to-image graph where a denoising_start value intended for img2img was left set.

Common situations: Reusing an img2img graph for txt2img without resetting denoising_start; a scheduler/denoise param node still feeding a nonzero start; saving denoising_start in a preset and switching modes.

Related errors

• denoising_start ({self.denoising_start}) must be less than d
• denoising_start should be 0 when initial latents are not pro
• No external provider config fields provided
• str(e)
• str(e) (ValueError from user service update, e.g. LastAdmini

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/1aa65cfefb8f4466. Report an issue: GitHub.