Help on function generate_image in module stable_diffusion_cpp.stable_diffusion:

generate_image(
    self,
    prompt: str,
    negative_prompt: str = '',
    clip_skip: int = -1,
    init_image: Union[PIL.Image.Image, str, NoneType] = None,
    ref_images: Optional[List[Union[PIL.Image.Image, str]]] = None,
    auto_resize_ref_image: bool = True,
    increase_ref_index: bool = False,
    mask_image: Union[PIL.Image.Image, str, NoneType] = None,
    width: int = 512,
    height: int = 512,
    cfg_scale: float = 7.0,
    image_cfg_scale: Optional[float] = None,
    guidance: float = 3.5,
    scheduler: Union[str, stable_diffusion_cpp.stable_diffusion_cpp.Scheduler, int, float, NoneType] = 'default',
    sample_method: Union[str, stable_diffusion_cpp.stable_diffusion_cpp.SampleMethod, int, float, NoneType] = 'default',
    sample_steps: int = 20,
    eta: float = 0.0,
    timestep_shift: int = 0,
    sigmas: Optional[str] = None,
    skip_layers: List[int] = [7, 8, 9],
    skip_layer_start: float = 0.01,
    skip_layer_end: float = 0.2,
    slg_scale: float = 0.0,
    strength: float = 0.75,
    seed: int = 42,
    batch_count: int = 1,
    control_image: Union[PIL.Image.Image, str, NoneType] = None,
    control_strength: float = 0.9,
    pm_id_embed_path: str = '',
    pm_id_images: Optional[List[Union[PIL.Image.Image, str]]] = None,
    pm_style_strength: float = 20.0,
    vae_tiling: bool = False,
    vae_tile_overlap: float = 0.5,
    vae_tile_size: Union[int, str, NoneType] = '0x0',
    vae_relative_tile_size: Union[float, str, NoneType] = '0x0',
    cache_mode: Union[str, stable_diffusion_cpp.stable_diffusion_cpp.SDCacheMode, int, float, NoneType] = 'disabled',
    cache_reuse_threshold: float = 1.0,
    cache_start_percent: float = 0.15,
    cache_end_percent: float = 0.95,
    cache_error_decay_rate: float = 1.0,
    cache_use_relative_threshold: bool = True,
    cache_reset_error_on_compute: bool = True,
    cache_Fn_compute_blocks: int = 8,
    cache_Bn_compute_blocks: int = 0,
    cache_residual_diff_threshold: float = 0.08,
    cache_max_warmup_steps: int = 8,
    cache_max_continuous_cached_steps: int = -1,
    cache_taylorseer_n_derivatives: int = 1,
    cache_taylorseer_skip_interval: int = 1,
    scm_mask: str = '',
    scm_policy: Literal['dynamic', 'static'] = 'dynamic',
    canny: bool = False,
    upscale_factor: int = 1,
    preview_method: Union[str, stable_diffusion_cpp.stable_diffusion_cpp.Preview, int, float] = 'none',
    preview_noisy: bool = False,
    preview_interval: int = 1,
    preview_callback: Optional[Callable] = None,
    progress_callback: Optional[Callable] = None
) -> List[PIL.Image.Image]
    Generate images from a text prompt and or input images.

    Args:
        prompt: The prompt to render.
        negative_prompt: The negative prompt.
        clip_skip: Ignore last layers of CLIP network (1 ignores none, 2 ignores one layer, <= 0 represents unspecified, will be 1 for SD1.x, 2 for SD2.x).
        init_image: An input image path or Pillow Image to direct the generation.
        ref_images: A list of input image paths or Pillow Images for Flux Kontext models (can be used multiple times).
        auto_resize_ref_image: Automatically resize reference images.
        increase_ref_index: Automatically increase the indices of reference images based on the order they are listed (starting with 1).
        mask_image: The inpainting mask image path or Pillow Image.
        width: Image width, in pixel space.
        height: Image height, in pixel space.
        cfg_scale: Unconditional guidance scale.
        image_cfg_scale: Image guidance scale for inpaint or instruct-pix2pix models.
        guidance: Distilled guidance scale for models with guidance input.
        scheduler: Denoiser sigma scheduler (default: discrete).
        sample_method: Sampling method (default: euler for Flux/SD3/Wan, euler_a otherwise).
        sample_steps: Number of sample steps.
        eta: Eta in DDIM, only for DDIM and TCD.
        timestep_shift: Shift timestep for NitroFusion models, default: 0, recommended N for NitroSD-Realism around 250 and 500 for NitroSD-Vibrant.
        sigmas: Custom sigma values for the sampler, comma-separated (e.g. "14.61,7.8,3.5,0.0").
        skip_layers: Layers to skip for SLG steps (SLG will be enabled at step int([STEPS]x[START]) and disabled at int([STEPS]x[END])).
        skip_layer_start: SLG enabling point.
        skip_layer_end: SLG disabling point.
        slg_scale: Skip layer guidance (SLG) scale, only for DiT models.
        strength: Strength for noising/unnoising.
        seed: RNG seed (uses random seed for < 0).
        batch_count: Number of images to generate.
        control_image: A control condition image path or Pillow Image (Control Net).
        control_strength: Strength to apply Control Net.
        pm_id_embed_path: Path to PhotoMaker v2 id embed.
        pm_id_images: A list of input image paths or Pillow Images for PhotoMaker input identity.
        pm_style_strength: Strength for keeping PhotoMaker input identity.
        vae_tiling: Process vae in tiles to reduce memory usage.
        vae_tile_overlap: Tile overlap for vae tiling, in fraction of tile size.
        vae_tile_size: Tile size for vae tiling ([X]x[Y] format).
        vae_relative_tile_size: Relative tile size for vae tiling, in fraction of image size if < 1, in number of tiles per dim if >=1 ([X]x[Y] format) (overrides `vae_tile_size`).
        cache_mode: The caching method to use (default: disabled).
        scm_mask: SCM steps mask for cache-dit: comma-separated 0/1 (e.g., "1,1,1,0,0,1,0,0,1,0") - 1=compute, 0=can cache.
        scm_policy: SCM policy 'dynamic' or 'static'.
        canny: Apply canny edge detection preprocessor to the `control_image`.
        upscale_factor: Run the ESRGAN upscaler this many times.
        preview_method: The preview method to use (default: none).
        preview_noisy: Enables previewing noisy inputs of the models rather than the denoised outputs.
        preview_interval: Interval in denoising steps between consecutive updates of the image preview (default: 1, meaning update at every step)
        preview_callback: Callback function to call on each preview frame.
        progress_callback: Callback function to call on each step end.

    Returns:
        A list of Pillow Images.

