pipeline_animatediff_video2video.py

# Copyright 2024 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import inspect
from typing import Any, Callable, Dict, List, Optional, Union

import torch
from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection

from ...image_processor import PipelineImageInput
from ...loaders import IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel, UNetMotionModel
from ...models.lora import adjust_lora_scale_text_encoder
from ...models.unets.unet_motion_model import MotionAdapter
from ...schedulers import (
    DDIMScheduler,
    DPMSolverMultistepScheduler,
    EulerAncestralDiscreteScheduler,
    EulerDiscreteScheduler,
    LMSDiscreteScheduler,
    PNDMScheduler,
)
from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers, unscale_lora_layers
from ...utils.torch_utils import randn_tensor
from ...video_processor import VideoProcessor
from ..free_init_utils import FreeInitMixin
from ..free_noise_utils import AnimateDiffFreeNoiseMixin
from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
from .pipeline_output import AnimateDiffPipelineOutput


logger = logging.get_logger(__name__)  # pylint: disable=invalid-name

EXAMPLE_DOC_STRING = """
    Examples:
        ```py
        >>> import imageio
        >>> import requests
        >>> import torch
        >>> from diffusers import AnimateDiffVideoToVideoPipeline, DDIMScheduler, MotionAdapter
        >>> from diffusers.utils import export_to_gif
        >>> from io import BytesIO
        >>> from PIL import Image

        >>> adapter = MotionAdapter.from_pretrained(
        ...     "guoyww/animatediff-motion-adapter-v1-5-2", torch_dtype=torch.float16
        ... )
        >>> pipe = AnimateDiffVideoToVideoPipeline.from_pretrained(
        ...     "SG161222/Realistic_Vision_V5.1_noVAE", motion_adapter=adapter
        ... ).to("cuda")
        >>> pipe.scheduler = DDIMScheduler(
        ...     beta_schedule="linear", steps_offset=1, clip_sample=False, timespace_spacing="linspace"
        ... )


        >>> def load_video(file_path: str):
        ...     images = []

        ...     if file_path.startswith(("http://", "https://")):
        ...         # If the file_path is a URL
        ...         response = requests.get(file_path)
        ...         response.raise_for_status()
        ...         content = BytesIO(response.content)
        ...         vid = imageio.get_reader(content)
        ...     else:
        ...         # Assuming it's a local file path
        ...         vid = imageio.get_reader(file_path)

        ...     for frame in vid:
        ...         pil_image = Image.fromarray(frame)
        ...         images.append(pil_image)

        ...     return images


        >>> video = load_video(
        ...     "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/animatediff-vid2vid-input-1.gif"
        ... )
        >>> output = pipe(
        ...     video=video, prompt="panda playing a guitar, on a boat, in the ocean, high quality", strength=0.5
        ... )
        >>> frames = output.frames[0]
        >>> export_to_gif(frames, "animation.gif")
        ```
"""


# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
def retrieve_latents(
    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
):
    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
        return encoder_output.latent_dist.sample(generator)
    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
        return encoder_output.latent_dist.mode()
    elif hasattr(encoder_output, "latents"):
        return encoder_output.latents
    else:
        raise AttributeError("Could not access latents of provided encoder_output")


# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
def retrieve_timesteps(
    scheduler,
    num_inference_steps: Optional[int] = None,
    device: Optional[Union[str, torch.device]] = None,
    timesteps: Optional[List[int]] = None,
    sigmas: Optional[List[float]] = None,
    **kwargs,
):
    """
    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.

    Args:
        scheduler (`SchedulerMixin`):
            The scheduler to get timesteps from.
        num_inference_steps (`int`):
            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
            must be `None`.
        device (`str` or `torch.device`, *optional*):
            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
        timesteps (`List[int]`, *optional*):
            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
            `num_inference_steps` and `sigmas` must be `None`.
        sigmas (`List[float]`, *optional*):
            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
            `num_inference_steps` and `timesteps` must be `None`.

    Returns:
        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
        second element is the number of inference steps.
    """
    if timesteps is not None and sigmas is not None:
        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
    if timesteps is not None:
        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
        if not accepts_timesteps:
            raise ValueError(
                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
                f" timestep schedules. Please check whether you are using the correct scheduler."
            )
        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
        timesteps = scheduler.timesteps
        num_inference_steps = len(timesteps)
    elif sigmas is not None:
        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
        if not accept_sigmas:
            raise ValueError(
                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
                f" sigmas schedules. Please check whether you are using the correct scheduler."
            )
        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
        timesteps = scheduler.timesteps
        num_inference_steps = len(timesteps)
    else:
        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
        timesteps = scheduler.timesteps
    return timesteps, num_inference_steps


class AnimateDiffVideoToVideoPipeline(
    DiffusionPipeline,
    StableDiffusionMixin,
    TextualInversionLoaderMixin,
    IPAdapterMixin,
    StableDiffusionLoraLoaderMixin,
    FreeInitMixin,
    AnimateDiffFreeNoiseMixin,
):
    r"""
    Pipeline for video-to-video generation.

    This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
    implemented for all pipelines (downloading, saving, running on a particular device, etc.).

    The pipeline also inherits the following loading methods:
        - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
        - [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
        - [`~loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
        - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters

    Args:
        vae ([`AutoencoderKL`]):
            Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.
        text_encoder ([`CLIPTextModel`]):
            Frozen text-encoder ([clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14)).
        tokenizer (`CLIPTokenizer`):
            A [`~transformers.CLIPTokenizer`] to tokenize text.
        unet ([`UNet2DConditionModel`]):
            A [`UNet2DConditionModel`] used to create a UNetMotionModel to denoise the encoded video latents.
        motion_adapter ([`MotionAdapter`]):
            A [`MotionAdapter`] to be used in combination with `unet` to denoise the encoded video latents.
        scheduler ([`SchedulerMixin`]):
            A scheduler to be used in combination with `unet` to denoise the encoded image latents. Can be one of
            [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
    """

    model_cpu_offload_seq = "text_encoder->image_encoder->unet->vae"
    _optional_components = ["feature_extractor", "image_encoder", "motion_adapter"]
    _callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"]

    def __init__(
        self,
        vae: AutoencoderKL,
        text_encoder: CLIPTextModel,
        tokenizer: CLIPTokenizer,
        unet: UNet2DConditionModel,
        motion_adapter: MotionAdapter,
        scheduler: Union[
            DDIMScheduler,
            PNDMScheduler,
            LMSDiscreteScheduler,
            EulerDiscreteScheduler,
            EulerAncestralDiscreteScheduler,
            DPMSolverMultistepScheduler,
        ],
        feature_extractor: CLIPImageProcessor = None,
        image_encoder: CLIPVisionModelWithProjection = None,
    ):
        super().__init__()
        if isinstance(unet, UNet2DConditionModel):
            unet = UNetMotionModel.from_unet2d(unet, motion_adapter)

        self.register_modules(
            vae=vae,
            text_encoder=text_encoder,
            tokenizer=tokenizer,
            unet=unet,
            motion_adapter=motion_adapter,
            scheduler=scheduler,
            feature_extractor=feature_extractor,
            image_encoder=image_encoder,
        )
        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
        self.video_processor = VideoProcessor(vae_scale_factor=self.vae_scale_factor)

    def encode_prompt(
        self,
        prompt,
        device,
        num_images_per_prompt,
        do_classifier_free_guidance,
        negative_prompt=None,
        prompt_embeds: Optional[torch.Tensor] = None,
        negative_prompt_embeds: Optional[torch.Tensor] = None,
        lora_scale: Optional[float] = None,
        clip_skip: Optional[int] = None,
    ):
        r"""
        Encodes the prompt into text encoder hidden states.

        Args:
            prompt (`str` or `List[str]`, *optional*):
                prompt to be encoded
            device: (`torch.device`):
                torch device
            num_images_per_prompt (`int`):
                number of images that should be generated per prompt
            do_classifier_free_guidance (`bool`):
                whether to use classifier free guidance or not
            negative_prompt (`str` or `List[str]`, *optional*):
                The prompt or prompts not to guide the image generation. If not defined, one has to pass
                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
                less than `1`).
            prompt_embeds (`torch.Tensor`, *optional*):
                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
                provided, text embeddings will be generated from `prompt` input argument.
            negative_prompt_embeds (`torch.Tensor`, *optional*):
                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
                argument.
            lora_scale (`float`, *optional*):
                A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
            clip_skip (`int`, *optional*):
                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
                the output of the pre-final layer will be used for computing the prompt embeddings.
        """
        # set lora scale so that monkey patched LoRA
        # function of text encoder can correctly access it
        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
            self._lora_scale = lora_scale

            # dynamically adjust the LoRA scale
            if not USE_PEFT_BACKEND:
                adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
            else:
                scale_lora_layers(self.text_encoder, lora_scale)

        if prompt is not None and isinstance(prompt, (str, dict)):
            batch_size = 1
        elif prompt is not None and isinstance(prompt, list):
            batch_size = len(prompt)
        else:
            batch_size = prompt_embeds.shape[0]

        if prompt_embeds is None:
            # textual inversion: process multi-vector tokens if necessary
            if isinstance(self, TextualInversionLoaderMixin):
                prompt = self.maybe_convert_prompt(prompt, self.tokenizer)

            text_inputs = self.tokenizer(
                prompt,
                padding="max_length",
                max_length=self.tokenizer.model_max_length,
                truncation=True,
                return_tensors="pt",
            )
            text_input_ids = text_inputs.input_ids
            untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids

            if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
                text_input_ids, untruncated_ids
            ):
                removed_text = self.tokenizer.batch_decode(
                    untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1]
                )
                logger.warning(
                    "The following part of your input was truncated because CLIP can only handle sequences up to"
                    f" {self.tokenizer.model_max_length} tokens: {removed_text}"
                )

            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
                attention_mask = text_inputs.attention_mask.to(device)
            else:
                attention_mask = None

            if clip_skip is None:
                prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
                prompt_embeds = prompt_embeds[0]
            else:
                prompt_embeds = self.text_encoder(
                    text_input_ids.to(device), attention_mask=attention_mask, output_hidden_states=True
                )
                # Access the `hidden_states` first, that contains a tuple of
                # all the hidden states from the encoder layers. Then index into
                # the tuple to access the hidden states from the desired layer.
                prompt_embeds = prompt_embeds[-1][-(clip_skip + 1)]
                # We also need to apply the final LayerNorm here to not mess with the
                # representations. The `last_hidden_states` that we typically use for
                # obtaining the final prompt representations passes through the LayerNorm
                # layer.
                prompt_embeds = self.text_encoder.text_model.final_layer_norm(prompt_embeds)

        if self.text_encoder is not None:
            prompt_embeds_dtype = self.text_encoder.dtype
        elif self.unet is not None:
            prompt_embeds_dtype = self.unet.dtype
        else:
            prompt_embeds_dtype = prompt_embeds.dtype

        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)

        bs_embed, seq_len, _ = prompt_embeds.shape
        # duplicate text embeddings for each generation per prompt, using mps friendly method
        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)

        # get unconditional embeddings for classifier free guidance
        if do_classifier_free_guidance and negative_prompt_embeds is None:
            uncond_tokens: List[str]
            if negative_prompt is None:
                uncond_tokens = [""] * batch_size
            elif prompt is not None and type(prompt) is not type(negative_prompt):
                raise TypeError(
                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
                    f" {type(prompt)}."
                )
            elif isinstance(negative_prompt, str):
                uncond_tokens = [negative_prompt]
            elif batch_size != len(negative_prompt):
                raise ValueError(
                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
                    " the batch size of `prompt`."
                )
            else:
                uncond_tokens = negative_prompt

            # textual inversion: process multi-vector tokens if necessary
            if isinstance(self, TextualInversionLoaderMixin):
                uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)

            max_length = prompt_embeds.shape[1]
            uncond_input = self.tokenizer(
                uncond_tokens,
                padding="max_length",
                max_length=max_length,
                truncation=True,
                return_tensors="pt",
            )

            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
                attention_mask = uncond_input.attention_mask.to(device)
            else:
                attention_mask = None

            negative_prompt_embeds = self.text_encoder(
                uncond_input.input_ids.to(device),
                attention_mask=attention_mask,
            )
            negative_prompt_embeds = negative_prompt_embeds[0]

        if do_classifier_free_guidance:
            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
            seq_len = negative_prompt_embeds.shape[1]

            negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)

            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)

        if self.text_encoder is not None:
            if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
                # Retrieve the original scale by scaling back the LoRA layers
                unscale_lora_layers(self.text_encoder, lora_scale)

        return prompt_embeds, negative_prompt_embeds

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_image
    def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):
        dtype = next(self.image_encoder.parameters()).dtype

        if not isinstance(image, torch.Tensor):
            image = self.feature_extractor(image, return_tensors="pt").pixel_values

        image = image.to(device=device, dtype=dtype)
        if output_hidden_states:
            image_enc_hidden_states = self.image_encoder(image, output_hidden_states=True).hidden_states[-2]
            image_enc_hidden_states = image_enc_hidden_states.repeat_interleave(num_images_per_prompt, dim=0)
            uncond_image_enc_hidden_states = self.image_encoder(
                torch.zeros_like(image), output_hidden_states=True
            ).hidden_states[-2]
            uncond_image_enc_hidden_states = uncond_image_enc_hidden_states.repeat_interleave(
                num_images_per_prompt, dim=0
            )
            return image_enc_hidden_states, uncond_image_enc_hidden_states
        else:
            image_embeds = self.image_encoder(image).image_embeds
            image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)
            uncond_image_embeds = torch.zeros_like(image_embeds)

            return image_embeds, uncond_image_embeds

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
    def prepare_ip_adapter_image_embeds(
        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt, do_classifier_free_guidance
    ):
        image_embeds = []
        if do_classifier_free_guidance:
            negative_image_embeds = []
        if ip_adapter_image_embeds is None:
            if not isinstance(ip_adapter_image, list):
                ip_adapter_image = [ip_adapter_image]

            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
                raise ValueError(
                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
                )

            for single_ip_adapter_image, image_proj_layer in zip(
                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
            ):
                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
                single_image_embeds, single_negative_image_embeds = self.encode_image(
                    single_ip_adapter_image, device, 1, output_hidden_state
                )

                image_embeds.append(single_image_embeds[None, :])
                if do_classifier_free_guidance:
                    negative_image_embeds.append(single_negative_image_embeds[None, :])
        else:
            for single_image_embeds in ip_adapter_image_embeds:
                if do_classifier_free_guidance:
                    single_negative_image_embeds, single_image_embeds = single_image_embeds.chunk(2)
                    negative_image_embeds.append(single_negative_image_embeds)
                image_embeds.append(single_image_embeds)

        ip_adapter_image_embeds = []
        for i, single_image_embeds in enumerate(image_embeds):
            single_image_embeds = torch.cat([single_image_embeds] * num_images_per_prompt, dim=0)
            if do_classifier_free_guidance:
                single_negative_image_embeds = torch.cat([negative_image_embeds[i]] * num_images_per_prompt, dim=0)
                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds], dim=0)

            single_image_embeds = single_image_embeds.to(device=device)
            ip_adapter_image_embeds.append(single_image_embeds)

        return ip_adapter_image_embeds

    def encode_video(self, video, generator, decode_chunk_size: int = 16) -> torch.Tensor:
        latents = []
        for i in range(0, len(video), decode_chunk_size):
            batch_video = video[i : i + decode_chunk_size]
            batch_video = retrieve_latents(self.vae.encode(batch_video), generator=generator)
            latents.append(batch_video)
        return torch.cat(latents)

    # Copied from diffusers.pipelines.animatediff.pipeline_animatediff.AnimateDiffPipeline.decode_latents
    def decode_latents(self, latents, decode_chunk_size: int = 16):
        latents = 1 / self.vae.config.scaling_factor * latents

        batch_size, channels, num_frames, height, width = latents.shape
        latents = latents.permute(0, 2, 1, 3, 4).reshape(batch_size * num_frames, channels, height, width)

        video = []
        for i in range(0, latents.shape[0], decode_chunk_size):
            batch_latents = latents[i : i + decode_chunk_size]
            batch_latents = self.vae.decode(batch_latents).sample
            video.append(batch_latents)

        video = torch.cat(video)
        video = video[None, :].reshape((batch_size, num_frames, -1) + video.shape[2:]).permute(0, 2, 1, 3, 4)
        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
        video = video.float()
        return video

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
    def prepare_extra_step_kwargs(self, generator, eta):
        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
        # eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502
        # and should be between [0, 1]

        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
        extra_step_kwargs = {}
        if accepts_eta:
            extra_step_kwargs["eta"] = eta

        # check if the scheduler accepts generator
        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
        if accepts_generator:
            extra_step_kwargs["generator"] = generator
        return extra_step_kwargs

    def check_inputs(
        self,
        prompt,
        strength,
        height,
        width,
        video=None,
        latents=None,
        negative_prompt=None,
        prompt_embeds=None,
        negative_prompt_embeds=None,
        ip_adapter_image=None,
        ip_adapter_image_embeds=None,
        callback_on_step_end_tensor_inputs=None,
    ):
        if strength < 0 or strength > 1:
            raise ValueError(f"The value of strength should in [0.0, 1.0] but is {strength}")

        if height % 8 != 0 or width % 8 != 0:
            raise ValueError(f"`height` and `width` have to be divisible by 8 but are {height} and {width}.")

        if callback_on_step_end_tensor_inputs is not None and not all(
            k in self._callback_tensor_inputs for k in callback_on_step_end_tensor_inputs
        ):
            raise ValueError(
                f"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}"
            )

        if prompt is not None and prompt_embeds is not None:
            raise ValueError(
                f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
                " only forward one of the two."
            )
        elif prompt is None and prompt_embeds is None:
            raise ValueError(
                "Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined."
            )
        elif prompt is not None and not isinstance(prompt, (str, list, dict)):
            raise ValueError(f"`prompt` has to be of type `str`, `list` or `dict` but is {type(prompt)}")

        if negative_prompt is not None and negative_prompt_embeds is not None:
            raise ValueError(
                f"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:"
                f" {negative_prompt_embeds}. Please make sure to only forward one of the two."
            )

        if prompt_embeds is not None and negative_prompt_embeds is not None:
            if prompt_embeds.shape != negative_prompt_embeds.shape:
                raise ValueError(
                    "`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but"
                    f" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`"
                    f" {negative_prompt_embeds.shape}."
                )

        if video is not None and latents is not None:
            raise ValueError("Only one of `video` or `latents` should be provided")

        if ip_adapter_image is not None and ip_adapter_image_embeds is not None:
            raise ValueError(
                "Provide either `ip_adapter_image` or `ip_adapter_image_embeds`. Cannot leave both `ip_adapter_image` and `ip_adapter_image_embeds` defined."
            )

        if ip_adapter_image_embeds is not None:
            if not isinstance(ip_adapter_image_embeds, list):
                raise ValueError(
                    f"`ip_adapter_image_embeds` has to be of type `list` but is {type(ip_adapter_image_embeds)}"
                )
            elif ip_adapter_image_embeds[0].ndim not in [3, 4]:
                raise ValueError(
                    f"`ip_adapter_image_embeds` has to be a list of 3D or 4D tensors but is {ip_adapter_image_embeds[0].ndim}D"
                )

    def get_timesteps(self, num_inference_steps, timesteps, strength, device):
        # get the original timestep using init_timestep
        init_timestep = min(int(num_inference_steps * strength), num_inference_steps)

        t_start = max(num_inference_steps - init_timestep, 0)
        timesteps = timesteps[t_start * self.scheduler.order :]

        return timesteps, num_inference_steps - t_start

    def prepare_latents(
        self,
        video: Optional[torch.Tensor] = None,
        height: int = 64,
        width: int = 64,
        num_channels_latents: int = 4,
        batch_size: int = 1,
        timestep: Optional[int] = None,
        dtype: Optional[torch.dtype] = None,
        device: Optional[torch.device] = None,
        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
        latents: Optional[torch.Tensor] = None,
        decode_chunk_size: int = 16,
        add_noise: bool = False,
    ) -> torch.Tensor:
        num_frames = video.shape[1] if latents is None else latents.shape[2]
        shape = (
            batch_size,
            num_channels_latents,
            num_frames,
            height // self.vae_scale_factor,
            width // self.vae_scale_factor,
        )

        if isinstance(generator, list) and len(generator) != batch_size:
            raise ValueError(
                f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
                f" size of {batch_size}. Make sure the batch size matches the length of the generators."
            )

        if latents is None:
            # make sure the VAE is in float32 mode, as it overflows in float16
            if self.vae.config.force_upcast:
                video = video.float()
                self.vae.to(dtype=torch.float32)

            if isinstance(generator, list):
                if len(generator) != batch_size:
                    raise ValueError(
                        f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
                        f" size of {batch_size}. Make sure the batch size matches the length of the generators."
                    )

                init_latents = [
                    self.encode_video(video[i], generator[i], decode_chunk_size).unsqueeze(0)
                    for i in range(batch_size)
                ]
            else:
                init_latents = [self.encode_video(vid, generator, decode_chunk_size).unsqueeze(0) for vid in video]

            init_latents = torch.cat(init_latents, dim=0)

            # restore vae to original dtype
            if self.vae.config.force_upcast:
                self.vae.to(dtype)

            init_latents = init_latents.to(dtype)
            init_latents = self.vae.config.scaling_factor * init_latents

            if batch_size > init_latents.shape[0] and batch_size % init_latents.shape[0] == 0:
                # expand init_latents for batch_size
                error_message = (
                    f"You have passed {batch_size} text prompts (`prompt`), but only {init_latents.shape[0]} initial"
                    " images (`image`). Please make sure to update your script to pass as many initial images as text prompts"
                )
                raise ValueError(error_message)
            elif batch_size > init_latents.shape[0] and batch_size % init_latents.shape[0] != 0:
                raise ValueError(
                    f"Cannot duplicate `image` of batch size {init_latents.shape[0]} to {batch_size} text prompts."
                )
            else:
                init_latents = torch.cat([init_latents], dim=0)

            noise = randn_tensor(init_latents.shape, generator=generator, device=device, dtype=dtype)
            latents = self.scheduler.add_noise(init_latents, noise, timestep).permute(0, 2, 1, 3, 4)
        else:
            if shape != latents.shape:
                # [B, C, F, H, W]
                raise ValueError(f"`latents` expected to have {shape=}, but found {latents.shape=}")

            latents = latents.to(device, dtype=dtype)

            if add_noise:
                noise = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
                latents = self.scheduler.add_noise(latents, noise, timestep)

        return latents

    @property
    def guidance_scale(self):
        return self._guidance_scale

    @property
    def clip_skip(self):
        return self._clip_skip

    # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
    # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
    # corresponds to doing no classifier free guidance.
    @property
    def do_classifier_free_guidance(self):
        return self._guidance_scale > 1

    @property
    def cross_attention_kwargs(self):
        return self._cross_attention_kwargs

    @property
    def num_timesteps(self):
        return self._num_timesteps

    @property
    def interrupt(self):
        return self._interrupt

    @torch.no_grad()
    def __call__(
        self,
        video: List[List[PipelineImageInput]] = None,
        prompt: Optional[Union[str, List[str]]] = None,
        height: Optional[int] = None,
        width: Optional[int] = None,
        num_inference_steps: int = 50,
        enforce_inference_steps: bool = False,
        timesteps: Optional[List[int]] = None,
        sigmas: Optional[List[float]] = None,
        guidance_scale: float = 7.5,
        strength: float = 0.8,
        negative_prompt: Optional[Union[str, List[str]]] = None,
        num_videos_per_prompt: Optional[int] = 1,
        eta: float = 0.0,
        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
        latents: Optional[torch.Tensor] = None,
        prompt_embeds: Optional[torch.Tensor] = None,
        negative_prompt_embeds: Optional[torch.Tensor] = None,
        ip_adapter_image: Optional[PipelineImageInput] = None,
        ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
        output_type: Optional[str] = "pil",
        return_dict: bool = True,
        cross_attention_kwargs: Optional[Dict[str, Any]] = None,
        clip_skip: Optional[int] = None,
        callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
        callback_on_step_end_tensor_inputs: List[str] = ["latents"],
        decode_chunk_size: int = 16,
    ):
        r"""
        The call function to the pipeline for generation.

        Args:
            video (`List[PipelineImageInput]`):
                The input video to condition the generation on. Must be a list of images/frames of the video.
            prompt (`str` or `List[str]`, *optional*):
                The prompt or prompts to guide image generation. If not defined, you need to pass `prompt_embeds`.
            height (`int`, *optional*, defaults to `self.unet.config.sample_size * self.vae_scale_factor`):
                The height in pixels of the generated video.
            width (`int`, *optional*, defaults to `self.unet.config.sample_size * self.vae_scale_factor`):
                The width in pixels of the generated video.
            num_inference_steps (`int`, *optional*, defaults to 50):
                The number of denoising steps. More denoising steps usually lead to a higher quality videos at the
                expense of slower inference.
            timesteps (`List[int]`, *optional*):
                Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument
                in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is
                passed will be used. Must be in descending order.
            sigmas (`List[float]`, *optional*):
                Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in
                their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed
                will be used.
            strength (`float`, *optional*, defaults to 0.8):
                Higher strength leads to more differences between original video and generated video.
            guidance_scale (`float`, *optional*, defaults to 7.5):
                A higher guidance scale value encourages the model to generate images closely linked to the text
                `prompt` at the expense of lower image quality. Guidance scale is enabled when `guidance_scale > 1`.
            negative_prompt (`str` or `List[str]`, *optional*):
                The prompt or prompts to guide what to not include in image generation. If not defined, you need to
                pass `negative_prompt_embeds` instead. Ignored when not using guidance (`guidance_scale < 1`).
            eta (`float`, *optional*, defaults to 0.0):
                Corresponds to parameter eta (η) from the [DDIM](https://arxiv.org/abs/2010.02502) paper. Only applies
                to the [`~schedulers.DDIMScheduler`], and is ignored in other schedulers.
            generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
                A [`torch.Generator`](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make
                generation deterministic.
            latents (`torch.Tensor`, *optional*):
                Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for video
                generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                tensor is generated by sampling using the supplied random `generator`. Latents should be of shape
                `(batch_size, num_channel, num_frames, height, width)`.
            prompt_embeds (`torch.Tensor`, *optional*):
                Pre-generated text embeddings. Can be used to easily tweak text inputs (prompt weighting). If not
                provided, text embeddings are generated from the `prompt` input argument.
            negative_prompt_embeds (`torch.Tensor`, *optional*):
                Pre-generated negative text embeddings. Can be used to easily tweak text inputs (prompt weighting). If
                not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.
            ip_adapter_image: (`PipelineImageInput`, *optional*):
                Optional image input to work with IP Adapters.
            ip_adapter_image_embeds (`List[torch.Tensor]`, *optional*):
                Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of
                IP-adapters. Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. It should
                contain the negative image embedding if `do_classifier_free_guidance` is set to `True`. If not
                provided, embeddings are computed from the `ip_adapter_image` input argument.
            output_type (`str`, *optional*, defaults to `"pil"`):
                The output format of the generated video. Choose between `torch.Tensor`, `PIL.Image` or `np.array`.
            return_dict (`bool`, *optional*, defaults to `True`):
                Whether or not to return a [`AnimateDiffPipelineOutput`] instead of a plain tuple.
            cross_attention_kwargs (`dict`, *optional*):
                A kwargs dictionary that if specified is passed along to the [`AttentionProcessor`] as defined in
                [`self.processor`](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
            clip_skip (`int`, *optional*):
                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
                the output of the pre-final layer will be used for computing the prompt embeddings.
            callback_on_step_end (`Callable`, *optional*):
                A function that calls at the end of each denoising steps during the inference. The function is called
                with the following arguments: `callback_on_step_end(self: DiffusionPipeline, step: int, timestep: int,
                callback_kwargs: Dict)`. `callback_kwargs` will include a list of all tensors as specified by
                `callback_on_step_end_tensor_inputs`.
            callback_on_step_end_tensor_inputs (`List`, *optional*):
                The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list
                will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the
                `._callback_tensor_inputs` attribute of your pipeline class.
            decode_chunk_size (`int`, defaults to `16`):
                The number of frames to decode at a time when calling `decode_latents` method.

        Examples:

        Returns:
            [`pipelines.animatediff.pipeline_output.AnimateDiffPipelineOutput`] or `tuple`:
                If `return_dict` is `True`, [`pipelines.animatediff.pipeline_output.AnimateDiffPipelineOutput`] is
                returned, otherwise a `tuple` is returned where the first element is a list with the generated frames.
        """

        # 0. Default height and width to unet
        height = height or self.unet.config.sample_size * self.vae_scale_factor
        width = width or self.unet.config.sample_size * self.vae_scale_factor

        num_videos_per_prompt = 1

        # 1. Check inputs. Raise error if not correct
        self.check_inputs(
            prompt=prompt,
            strength=strength,
            height=height,
            width=width,
            negative_prompt=negative_prompt,
            prompt_embeds=prompt_embeds,
            negative_prompt_embeds=negative_prompt_embeds,
            video=video,
            latents=latents,
            ip_adapter_image=ip_adapter_image,
            ip_adapter_image_embeds=ip_adapter_image_embeds,
            callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs,
        )

        self._guidance_scale = guidance_scale
        self._clip_skip = clip_skip
        self._cross_attention_kwargs = cross_attention_kwargs
        self._interrupt = False

        # 2. Define call parameters
        if prompt is not None and isinstance(prompt, (str, dict)):
            batch_size = 1
        elif prompt is not None and isinstance(prompt, list):
            batch_size = len(prompt)
        else:
            batch_size = prompt_embeds.shape[0]

        device = self._execution_device
        dtype = self.dtype

        # 3. Prepare timesteps
        if not enforce_inference_steps:
            timesteps, num_inference_steps = retrieve_timesteps(
                self.scheduler, num_inference_steps, device, timesteps, sigmas
            )
            timesteps, num_inference_steps = self.get_timesteps(num_inference_steps, timesteps, strength, device)
            latent_timestep = timesteps[:1].repeat(batch_size * num_videos_per_prompt)
        else:
            denoising_inference_steps = int(num_inference_steps / strength)
            timesteps, denoising_inference_steps = retrieve_timesteps(
                self.scheduler, denoising_inference_steps, device, timesteps, sigmas
            )
            timesteps = timesteps[-num_inference_steps:]
            latent_timestep = timesteps[:1].repeat(batch_size * num_videos_per_prompt)

        # 4. Prepare latent variables
        if latents is None:
            video = self.video_processor.preprocess_video(video, height=height, width=width)
            # Move the number of frames before the number of channels.
            video = video.permute(0, 2, 1, 3, 4)
            video = video.to(device=device, dtype=dtype)
        num_channels_latents = self.unet.config.in_channels
        latents = self.prepare_latents(
            video=video,
            height=height,
            width=width,
            num_channels_latents=num_channels_latents,
            batch_size=batch_size * num_videos_per_prompt,
            timestep=latent_timestep,
            dtype=dtype,
            device=device,
            generator=generator,
            latents=latents,
            decode_chunk_size=decode_chunk_size,
            add_noise=enforce_inference_steps,
        )

        # 5. Encode input prompt
        text_encoder_lora_scale = (
            self.cross_attention_kwargs.get("scale", None) if self.cross_attention_kwargs is not None else None
        )
        num_frames = latents.shape[2]
        if self.free_noise_enabled:
            prompt_embeds, negative_prompt_embeds = self._encode_prompt_free_noise(
                prompt=prompt,
                num_frames=num_frames,
                device=device,
                num_videos_per_prompt=num_videos_per_prompt,
                do_classifier_free_guidance=self.do_classifier_free_guidance,
                negative_prompt=negative_prompt,
                prompt_embeds=prompt_embeds,
                negative_prompt_embeds=negative_prompt_embeds,
                lora_scale=text_encoder_lora_scale,
                clip_skip=self.clip_skip,
            )
        else:
            prompt_embeds, negative_prompt_embeds = self.encode_prompt(
                prompt,
                device,
                num_videos_per_prompt,
                self.do_classifier_free_guidance,
                negative_prompt,
                prompt_embeds=prompt_embeds,
                negative_prompt_embeds=negative_prompt_embeds,
                lora_scale=text_encoder_lora_scale,
                clip_skip=self.clip_skip,
            )

            # For classifier free guidance, we need to do two forward passes.
            # Here we concatenate the unconditional and text embeddings into a single batch
            # to avoid doing two forward passes
            if self.do_classifier_free_guidance:
                prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds])

            prompt_embeds = prompt_embeds.repeat_interleave(repeats=num_frames, dim=0)

        # 6. Prepare IP-Adapter embeddings
        if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
            image_embeds = self.prepare_ip_adapter_image_embeds(
                ip_adapter_image,
                ip_adapter_image_embeds,
                device,
                batch_size * num_videos_per_prompt,
                self.do_classifier_free_guidance,
            )

        # 7. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline
        extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)

        # 8. Add image embeds for IP-Adapter
        added_cond_kwargs = (
            {"image_embeds": image_embeds}
            if ip_adapter_image is not None or ip_adapter_image_embeds is not None
            else None
        )

        num_free_init_iters = self._free_init_num_iters if self.free_init_enabled else 1
        for free_init_iter in range(num_free_init_iters):
            if self.free_init_enabled:
                latents, timesteps = self._apply_free_init(
                    latents, free_init_iter, num_inference_steps, device, latents.dtype, generator
                )
                num_inference_steps = len(timesteps)
                # make sure to readjust timesteps based on strength
                timesteps, num_inference_steps = self.get_timesteps(num_inference_steps, timesteps, strength, device)

            self._num_timesteps = len(timesteps)
            num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order

            # 9. Denoising loop
            with self.progress_bar(total=self._num_timesteps) as progress_bar:
                for i, t in enumerate(timesteps):
                    if self.interrupt:
                        continue

                    # expand the latents if we are doing classifier free guidance
                    latent_model_input = torch.cat([latents] * 2) if self.do_classifier_free_guidance else latents
                    latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)

                    # predict the noise residual
                    noise_pred = self.unet(
                        latent_model_input,
                        t,
                        encoder_hidden_states=prompt_embeds,
                        cross_attention_kwargs=self.cross_attention_kwargs,
                        added_cond_kwargs=added_cond_kwargs,
                    ).sample

                    # perform guidance
                    if self.do_classifier_free_guidance:
                        noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
                        noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)

                    # compute the previous noisy sample x_t -> x_t-1
                    latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).prev_sample

                    if callback_on_step_end is not None:
                        callback_kwargs = {}
                        for k in callback_on_step_end_tensor_inputs:
                            callback_kwargs[k] = locals()[k]
                        callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)

                        latents = callback_outputs.pop("latents", latents)
                        prompt_embeds = callback_outputs.pop("prompt_embeds", prompt_embeds)
                        negative_prompt_embeds = callback_outputs.pop("negative_prompt_embeds", negative_prompt_embeds)

                    # call the callback, if provided
                    if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
                        progress_bar.update()

        # 10. Post-processing
        if output_type == "latent":
            video = latents
        else:
            video_tensor = self.decode_latents(latents, decode_chunk_size)
            video = self.video_processor.postprocess_video(video=video_tensor, output_type=output_type)

        # 11. Offload all models
        self.maybe_free_model_hooks()

        if not return_dict:
            return (video,)

        return AnimateDiffPipelineOutput(frames=video)