diffusers/pr_11660/en/_app/immutable/nodes/170.f6f01963.js

import{s as zt,o as St,n as Xe}from"../chunks/scheduler.8c3d61f6.js";import{S as Pt,i as At,g as p,s as i,r as y,A as Lt,h as c,f as s,c as r,j as I,x as f,u as h,k as Q,l as qt,y as o,a as u,v as g,d as b,t as J,w as _}from"../chunks/index.da70eac4.js";import{T as Dt}from"../chunks/Tip.1d9b8c37.js";import{D as ue}from"../chunks/Docstring.9419aa1d.js";import{C as ke}from"../chunks/CodeBlock.a9c4becf.js";import{E as Et}from"../chunks/ExampleCodeBlock.1b2603c3.js";import{H as Ce,E as Kt}from"../chunks/getInferenceSnippets.39110341.js";import{H as Ot,a as $t}from"../chunks/HfOption.6ab18950.js";function en(W){let n,w="Click on the Wan2.1 models in the right sidebar for more examples of video generation.";return{c(){n=p("p"),n.textContent=w},l(m){n=c(m,"P",{"data-svelte-h":!0}),f(n)!=="svelte-1jx6um4"&&(n.textContent=w)},m(m,d){u(m,n,d)},p:Xe,d(m){m&&s(n)}}}function tn(W){let n,w='Refer to the <a href="../../optimization/memory">Reduce memory usage</a> guide for more details about the various memory saving techniques.',m,d,l="The Wan2.1 text-to-video model below requires ~13GB of VRAM.",t,M,U;return M=new ke({props:{code:"JTIzJTIwcGlwJTIwaW5zdGFsbCUyMGZ0ZnklMEFpbXBvcnQlMjB0b3JjaCUwQWltcG9ydCUyMG51bXB5JTIwYXMlMjBucCUwQWZyb20lMjBkaWZmdXNlcnMlMjBpbXBvcnQlMjBBdXRvTW9kZWwlMkMlMjBXYW5QaXBlbGluZSUwQWZyb20lMjBkaWZmdXNlcnMucXVhbnRpemVycyUyMGltcG9ydCUyMFBpcGVsaW5lUXVhbnRpemF0aW9uQ29uZmlnJTBBZnJvbSUyMGRpZmZ1c2Vycy5ob29rcy5ncm91cF9vZmZsb2FkaW5nJTIwaW1wb3J0JTIwYXBwbHlfZ3JvdXBfb2ZmbG9hZGluZyUwQWZyb20lMjBkaWZmdXNlcnMudXRpbHMlMjBpbXBvcnQlMjBleHBvcnRfdG9fdmlkZW8lMkMlMjBsb2FkX2ltYWdlJTBBZnJvbSUyMHRyYW5zZm9ybWVycyUyMGltcG9ydCUyMFVNVDVFbmNvZGVyTW9kZWwlMEElMEF0ZXh0X2VuY29kZXIlMjAlM0QlMjBVTVQ1RW5jb2Rlck1vZGVsLmZyb21fcHJldHJhaW5lZCglMjJXYW4tQUklMkZXYW4yLjEtVDJWLTE0Qi1EaWZmdXNlcnMlMjIlMkMlMjBzdWJmb2xkZXIlM0QlMjJ0ZXh0X2VuY29kZXIlMjIlMkMlMjB0b3JjaF9kdHlwZSUzRHRvcmNoLmJmbG9hdDE2KSUwQXZhZSUyMCUzRCUyMEF1dG9Nb2RlbC5mcm9tX3ByZXRyYWluZWQoJTIyV2FuLUFJJTJGV2FuMi4xLVQyVi0xNEItRGlmZnVzZXJzJTIyJTJDJTIwc3ViZm9sZGVyJTNEJTIydmFlJTIyJTJDJTIwdG9yY2hfZHR5cGUlM0R0b3JjaC5mbG9hdDMyKSUwQXRyYW5zZm9ybWVyJTIwJTNEJTIwQXV0b01vZGVsLmZyb21fcHJldHJhaW5lZCglMjJXYW4tQUklMkZXYW4yLjEtVDJWLTE0Qi1EaWZmdXNlcnMlMjIlMkMlMjBzdWJmb2xkZXIlM0QlMjJ0cmFuc2Zvcm1lciUyMiUyQyUyMHRvcmNoX2R0eXBlJTNEdG9yY2guYmZsb2F0MTYpJTBBJTBBJTIzJTIwZ3JvdXAtb2ZmbG9hZGluZyUwQW9ubG9hZF9kZXZpY2UlMjAlM0QlMjB0b3JjaC5kZXZpY2UoJTIyY3VkYSUyMiklMEFvZmZsb2FkX2RldmljZSUyMCUzRCUyMHRvcmNoLmRldmljZSglMjJjcHUlMjIpJTBBYXBwbHlfZ3JvdXBfb2ZmbG9hZGluZyh0ZXh0X2VuY29kZXIlMkMlMEElMjAlMjAlMjAlMjBvbmxvYWRfZGV2aWNlJTNEb25sb2FkX2RldmljZSUyQyUwQSUyMCUyMCUyMCUyMG9mZmxvYWRfZGV2aWNlJTNEb2ZmbG9hZF9kZXZpY2UlMkMlMEElMjAlMjAlMjAlMjBvZmZsb2FkX3R5cGUlM0QlMjJibG9ja19sZXZlbCUyMiUyQyUwQSUyMCUyMCUyMCUyMG51bV9ibG9ja3NfcGVyX2dyb3VwJTNENCUwQSklMEF0cmFuc2Zvcm1lci5lbmFibGVfZ3JvdXBfb2ZmbG9hZCglMEElMjAlMjAlMjAlMjBvbmxvYWRfZGV2aWNlJTNEb25sb2FkX2RldmljZSUyQyUwQSUyMCUyMCUyMCUyMG9mZmxvYWRfZGV2aWNlJTNEb2ZmbG9hZF9kZXZpY2UlMkMlMEElMjAlMjAlMjAlMjBvZmZsb2FkX3R5cGUlM0QlMjJsZWFmX2xldmVsJTIyJTJDJTBBJTIwJTIwJTIwJTIwdXNlX3N0cmVhbSUzRFRydWUlMEEpJTBBJTBBcGlwZWxpbmUlMjAlM0QlMjBXYW5QaXBlbGluZS5mcm9tX3ByZXRyYWluZWQoJTBBJTIwJTIwJTIwJTIwJTIyV2FuLUFJJTJGV2FuMi4xLVQyVi0xNEItRGlmZnVzZXJzJTIyJTJDJTBBJTIwJTIwJTIwJTIwdmFlJTNEdmFlJTJDJTBBJTIwJTIwJTIwJTIwdHJhbnNmb3JtZXIlM0R0cmFuc2Zvcm1lciUyQyUwQSUyMCUyMCUyMCUyMHRleHRfZW5jb2RlciUzRHRleHRfZW5jb2RlciUyQyUwQSUyMCUyMCUyMCUyMHRvcmNoX2R0eXBlJTNEdG9yY2guYmZsb2F0MTYlMEEpJTBBcGlwZWxpbmUudG8oJTIyY3VkYSUyMiklMEElMEFwcm9tcHQlMjAlM0QlMjAlMjIlMjIlMjIlMEFUaGUlMjBjYW1lcmElMjBydXNoZXMlMjBmcm9tJTIwZmFyJTIwdG8lMjBuZWFyJTIwaW4lMjBhJTIwbG93LWFuZ2xlJTIwc2hvdCUyQyUyMCUwQXJldmVhbGluZyUyMGElMjB3aGl0ZSUyMGZlcnJldCUyMG9uJTIwYSUyMGxvZy4lMjBJdCUyMHBsYXlzJTJDJTIwbGVhcHMlMjBpbnRvJTIwdGhlJTIwd2F0ZXIlMkMlMjBhbmQlMjBlbWVyZ2VzJTJDJTIwYXMlMjB0aGUlMjBjYW1lcmElMjB6b29tcyUyMGluJTIwJTBBZm9yJTIwYSUyMGNsb3NlLXVwLiUyMFdhdGVyJTIwc3BsYXNoZXMlMjBiZXJyeSUyMGJ1c2hlcyUyMG5lYXJieSUyQyUyMHdoaWxlJTIwbW9zcyUyQyUyMHNub3clMkMlMjBhbmQlMjBsZWF2ZXMlMjBibGFua2V0JTIwdGhlJTIwZ3JvdW5kLiUyMCUwQUJpcmNoJTIwdHJlZXMlMjBhbmQlMjBhJTIwbGlnaHQlMjBibHVlJTIwc2t5JTIwZnJhbWUlMjB0aGUlMjBzY2VuZSUyQyUyMHdpdGglMjBmZXJucyUyMGluJTIwdGhlJTIwZm9yZWdyb3VuZC4lMjBTaWRlJTIwbGlnaHRpbmclMjBjYXN0cyUyMGR5bmFtaWMlMjAlMEFzaGFkb3dzJTIwYW5kJTIwd2FybSUyMGhpZ2hsaWdodHMuJTIwTWVkaXVtJTIwY29tcG9zaXRpb24lMkMlMjBmcm9udCUyMHZpZXclMkMlMjBsb3clMjBhbmdsZSUyQyUyMHdpdGglMjBkZXB0aCUyMG9mJTIwZmllbGQuJTBBJTIyJTIyJTIyJTBBbmVnYXRpdmVfcHJvbXB0JTIwJTNEJTIwJTIyJTIyJTIyJTBBQnJpZ2h0JTIwdG9uZXMlMkMlMjBvdmVyZXhwb3NlZCUyQyUyMHN0YXRpYyUyQyUyMGJsdXJyZWQlMjBkZXRhaWxzJTJDJTIwc3VidGl0bGVzJTJDJTIwc3R5bGUlMkMlMjB3b3JrcyUyQyUyMHBhaW50aW5ncyUyQyUyMGltYWdlcyUyQyUyMHN0YXRpYyUyQyUyMG92ZXJhbGwlMjBncmF5JTJDJTIwd29yc3QlMjBxdWFsaXR5JTJDJTIwJTBBbG93JTIwcXVhbGl0eSUyQyUyMEpQRUclMjBjb21wcmVzc2lvbiUyMHJlc2lkdWUlMkMlMjB1Z2x5JTJDJTIwaW5jb21wbGV0ZSUyQyUyMGV4dHJhJTIwZmluZ2VycyUyQyUyMHBvb3JseSUyMGRyYXduJTIwaGFuZHMlMkMlMjBwb29ybHklMjBkcmF3biUyMGZhY2VzJTJDJTIwZGVmb3JtZWQlMkMlMjBkaXNmaWd1cmVkJTJDJTIwJTBBbWlzc2hhcGVuJTIwbGltYnMlMkMlMjBmdXNlZCUyMGZpbmdlcnMlMkMlMjBzdGlsbCUyMHBpY3R1cmUlMkMlMjBtZXNzeSUyMGJhY2tncm91bmQlMkMlMjB0aHJlZSUyMGxlZ3MlMkMlMjBtYW55JTIwcGVvcGxlJTIwaW4lMjB0aGUlMjBiYWNrZ3JvdW5kJTJDJTIwd2Fsa2luZyUyMGJhY2t3YXJkcyUwQSUyMiUyMiUyMiUwQSUwQW91dHB1dCUyMCUzRCUyMHBpcGVsaW5lKCUwQSUyMCUyMCUyMCUyMHByb21wdCUzRHByb21wdCUyQyUwQSUyMCUyMCUyMCUyMG5lZ2F0aXZlX3Byb21wdCUzRG5lZ2F0aXZlX3Byb21wdCUyQyUwQSUyMCUyMCUyMCUyMG51bV9mcmFtZXMlM0Q4MSUyQyUwQSUyMCUyMCUyMCUyMGd1aWRhbmNlX3NjYWxlJTNENS4wJTJDJTBBKS5mcmFtZXMlNUIwJTVEJTBBZXhwb3J0X3RvX3ZpZGVvKG91dHB1dCUyQyUyMCUyMm91dHB1dC5tcDQlMjIlMkMlMjBmcHMlM0QxNik=",highlighted:`<span class="hljs-comment"># pip install ftfy</span>
<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">import</span> numpy <span class="hljs-keyword">as</span> np
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel, WanPipeline
<span class="hljs-keyword">from</span> diffusers.quantizers <span class="hljs-keyword">import</span> PipelineQuantizationConfig
<span class="hljs-keyword">from</span> diffusers.hooks.group_offloading <span class="hljs-keyword">import</span> apply_group_offloading
<span class="hljs-keyword">from</span> diffusers.utils <span class="hljs-keyword">import</span> export_to_video, load_image
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> UMT5EncoderModel

text_encoder = UMT5EncoderModel.from_pretrained(<span class="hljs-string">&quot;Wan-AI/Wan2.1-T2V-14B-Diffusers&quot;</span>, subfolder=<span class="hljs-string">&quot;text_encoder&quot;</span>, torch_dtype=torch.bfloat16)
vae = AutoModel.from_pretrained(<span class="hljs-string">&quot;Wan-AI/Wan2.1-T2V-14B-Diffusers&quot;</span>, subfolder=<span class="hljs-string">&quot;vae&quot;</span>, torch_dtype=torch.float32)
transformer = AutoModel.from_pretrained(<span class="hljs-string">&quot;Wan-AI/Wan2.1-T2V-14B-Diffusers&quot;</span>, subfolder=<span class="hljs-string">&quot;transformer&quot;</span>, torch_dtype=torch.bfloat16)

<span class="hljs-comment"># group-offloading</span>
onload_device = torch.device(<span class="hljs-string">&quot;cuda&quot;</span>)
offload_device = torch.device(<span class="hljs-string">&quot;cpu&quot;</span>)
apply_group_offloading(text_encoder,
    onload_device=onload_device,
    offload_device=offload_device,
    offload_type=<span class="hljs-string">&quot;block_level&quot;</span>,
    num_blocks_per_group=<span class="hljs-number">4</span>
)
transformer.enable_group_offload(
    onload_device=onload_device,
    offload_device=offload_device,
    offload_type=<span class="hljs-string">&quot;leaf_level&quot;</span>,
    use_stream=<span class="hljs-literal">True</span>
)

pipeline = WanPipeline.from_pretrained(
    <span class="hljs-string">&quot;Wan-AI/Wan2.1-T2V-14B-Diffusers&quot;</span>,
    vae=vae,
    transformer=transformer,
    text_encoder=text_encoder,
    torch_dtype=torch.bfloat16
)
pipeline.to(<span class="hljs-string">&quot;cuda&quot;</span>)

prompt = <span class="hljs-string">&quot;&quot;&quot;
The camera rushes from far to near in a low-angle shot, 
revealing a white ferret on a log. It plays, leaps into the water, and emerges, as the camera zooms in 
for a close-up. Water splashes berry bushes nearby, while moss, snow, and leaves blanket the ground. 
Birch trees and a light blue sky frame the scene, with ferns in the foreground. Side lighting casts dynamic 
shadows and warm highlights. Medium composition, front view, low angle, with depth of field.
&quot;&quot;&quot;</span>
negative_prompt = <span class="hljs-string">&quot;&quot;&quot;
Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, 
low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, 
misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards
&quot;&quot;&quot;</span>

output = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    num_frames=<span class="hljs-number">81</span>,
    guidance_scale=<span class="hljs-number">5.0</span>,
).frames[<span class="hljs-number">0</span>]
export_to_video(output, <span class="hljs-string">&quot;output.mp4&quot;</span>, fps=<span class="hljs-number">16</span>)`,wrap:!1}}),{c(){n=p("p"),n.innerHTML=w,m=i(),d=p("p"),d.textContent=l,t=i(),y(M.$$.fragment)},l(T){n=c(T,"P",{"data-svelte-h":!0}),f(n)!=="svelte-iowzkr"&&(n.innerHTML=w),m=r(T),d=c(T,"P",{"data-svelte-h":!0}),f(d)!=="svelte-lj5f0r"&&(d.textContent=l),t=r(T),h(M.$$.fragment,T)},m(T,B){u(T,n,B),u(T,m,B),u(T,d,B),u(T,t,B),g(M,T,B),U=!0},p:Xe,i(T){U||(b(M.$$.fragment,T),U=!0)},o(T){J(M.$$.fragment,T),U=!1},d(T){T&&(s(n),s(m),s(d),s(t)),_(M,T)}}}function nn(W){let n,w='<a href="../../optimization/fp16#torchcompile">Compilation</a> is slow the first time but subsequent calls to the pipeline are faster.',m,d,l;return d=new ke({props:{code:"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",highlighted:`<span class="hljs-comment"># pip install ftfy</span>
<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">import</span> numpy <span class="hljs-keyword">as</span> np
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel, WanPipeline
<span class="hljs-keyword">from</span> diffusers.hooks.group_offloading <span class="hljs-keyword">import</span> apply_group_offloading
<span class="hljs-keyword">from</span> diffusers.utils <span class="hljs-keyword">import</span> export_to_video, load_image
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> UMT5EncoderModel

text_encoder = UMT5EncoderModel.from_pretrained(<span class="hljs-string">&quot;Wan-AI/Wan2.1-T2V-14B-Diffusers&quot;</span>, subfolder=<span class="hljs-string">&quot;text_encoder&quot;</span>, torch_dtype=torch.bfloat16)
vae = AutoModel.from_pretrained(<span class="hljs-string">&quot;Wan-AI/Wan2.1-T2V-14B-Diffusers&quot;</span>, subfolder=<span class="hljs-string">&quot;vae&quot;</span>, torch_dtype=torch.float32)
transformer = AutoModel.from_pretrained(<span class="hljs-string">&quot;Wan-AI/Wan2.1-T2V-14B-Diffusers&quot;</span>, subfolder=<span class="hljs-string">&quot;transformer&quot;</span>, torch_dtype=torch.bfloat16)

pipeline = WanPipeline.from_pretrained(
    <span class="hljs-string">&quot;Wan-AI/Wan2.1-T2V-14B-Diffusers&quot;</span>,
    vae=vae,
    transformer=transformer,
    text_encoder=text_encoder,
    torch_dtype=torch.bfloat16
)
pipeline.to(<span class="hljs-string">&quot;cuda&quot;</span>)

<span class="hljs-comment"># torch.compile</span>
pipeline.transformer.to(memory_format=torch.channels_last)
pipeline.transformer = torch.<span class="hljs-built_in">compile</span>(
    pipeline.transformer, mode=<span class="hljs-string">&quot;max-autotune&quot;</span>, fullgraph=<span class="hljs-literal">True</span>
)

prompt = <span class="hljs-string">&quot;&quot;&quot;
The camera rushes from far to near in a low-angle shot, 
revealing a white ferret on a log. It plays, leaps into the water, and emerges, as the camera zooms in 
for a close-up. Water splashes berry bushes nearby, while moss, snow, and leaves blanket the ground. 
Birch trees and a light blue sky frame the scene, with ferns in the foreground. Side lighting casts dynamic 
shadows and warm highlights. Medium composition, front view, low angle, with depth of field.
&quot;&quot;&quot;</span>
negative_prompt = <span class="hljs-string">&quot;&quot;&quot;
Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, 
low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, 
misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards
&quot;&quot;&quot;</span>

output = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    num_frames=<span class="hljs-number">81</span>,
    guidance_scale=<span class="hljs-number">5.0</span>,
).frames[<span class="hljs-number">0</span>]
export_to_video(output, <span class="hljs-string">&quot;output.mp4&quot;</span>, fps=<span class="hljs-number">16</span>)`,wrap:!1}}),{c(){n=p("p"),n.innerHTML=w,m=i(),y(d.$$.fragment)},l(t){n=c(t,"P",{"data-svelte-h":!0}),f(n)!=="svelte-dcc01q"&&(n.innerHTML=w),m=r(t),h(d.$$.fragment,t)},m(t,M){u(t,n,M),u(t,m,M),g(d,t,M),l=!0},p:Xe,i(t){l||(b(d.$$.fragment,t),l=!0)},o(t){J(d.$$.fragment,t),l=!1},d(t){t&&(s(n),s(m)),_(d,t)}}}function sn(W){let n,w,m,d;return n=new $t({props:{id:"usage",option:"memory",$$slots:{default:[tn]},$$scope:{ctx:W}}}),m=new $t({props:{id:"usage",option:"inference speed",$$slots:{default:[nn]},$$scope:{ctx:W}}}),{c(){y(n.$$.fragment),w=i(),y(m.$$.fragment)},l(l){h(n.$$.fragment,l),w=r(l),h(m.$$.fragment,l)},m(l,t){g(n,l,t),u(l,w,t),g(m,l,t),d=!0},p(l,t){const M={};t&2&&(M.$$scope={dirty:t,ctx:l}),n.$set(M);const U={};t&2&&(U.$$scope={dirty:t,ctx:l}),m.$set(U)},i(l){d||(b(n.$$.fragment,l),b(m.$$.fragment,l),d=!0)},o(l){J(n.$$.fragment,l),J(m.$$.fragment,l),d=!1},d(l){l&&s(w),_(n,l),_(m,l)}}}function an(W){let n,w="Examples:",m,d,l;return d=new ke({props:{code:"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",highlighted:`<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">import</span> torch
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> diffusers.utils <span class="hljs-keyword">import</span> export_to_video
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoencoderKLWan, WanPipeline
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> diffusers.schedulers.scheduling_unipc_multistep <span class="hljs-keyword">import</span> UniPCMultistepScheduler

<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-comment"># Available models: Wan-AI/Wan2.1-T2V-14B-Diffusers, Wan-AI/Wan2.1-T2V-1.3B-Diffusers</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>model_id = <span class="hljs-string">&quot;Wan-AI/Wan2.1-T2V-14B-Diffusers&quot;</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>vae = AutoencoderKLWan.from_pretrained(model_id, subfolder=<span class="hljs-string">&quot;vae&quot;</span>, torch_dtype=torch.float32)
<span class="hljs-meta">&gt;&gt;&gt; </span>pipe = WanPipeline.from_pretrained(model_id, vae=vae, torch_dtype=torch.bfloat16)
<span class="hljs-meta">&gt;&gt;&gt; </span>flow_shift = <span class="hljs-number">5.0</span>  <span class="hljs-comment"># 5.0 for 720P, 3.0 for 480P</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config, flow_shift=flow_shift)
<span class="hljs-meta">&gt;&gt;&gt; </span>pipe.to(<span class="hljs-string">&quot;cuda&quot;</span>)

<span class="hljs-meta">&gt;&gt;&gt; </span>prompt = <span class="hljs-string">&quot;A cat and a dog baking a cake together in a kitchen. The cat is carefully measuring flour, while the dog is stirring the batter with a wooden spoon. The kitchen is cozy, with sunlight streaming through the window.&quot;</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>negative_prompt = <span class="hljs-string">&quot;Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards&quot;</span>

<span class="hljs-meta">&gt;&gt;&gt; </span>output = pipe(
<span class="hljs-meta">... </span>    prompt=prompt,
<span class="hljs-meta">... </span>    negative_prompt=negative_prompt,
<span class="hljs-meta">... </span>    height=<span class="hljs-number">720</span>,
<span class="hljs-meta">... </span>    width=<span class="hljs-number">1280</span>,
<span class="hljs-meta">... </span>    num_frames=<span class="hljs-number">81</span>,
<span class="hljs-meta">... </span>    guidance_scale=<span class="hljs-number">5.0</span>,
<span class="hljs-meta">... </span>).frames[<span class="hljs-number">0</span>]
<span class="hljs-meta">&gt;&gt;&gt; </span>export_to_video(output, <span class="hljs-string">&quot;output.mp4&quot;</span>, fps=<span class="hljs-number">16</span>)`,wrap:!1}}),{c(){n=p("p"),n.textContent=w,m=i(),y(d.$$.fragment)},l(t){n=c(t,"P",{"data-svelte-h":!0}),f(n)!=="svelte-kvfsh7"&&(n.textContent=w),m=r(t),h(d.$$.fragment,t)},m(t,M){u(t,n,M),u(t,m,M),g(d,t,M),l=!0},p:Xe,i(t){l||(b(d.$$.fragment,t),l=!0)},o(t){J(d.$$.fragment,t),l=!1},d(t){t&&(s(n),s(m)),_(d,t)}}}function on(W){let n,w="Examples:",m,d,l;return d=new ke({props:{code:"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",highlighted:`<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">import</span> torch
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">import</span> numpy <span class="hljs-keyword">as</span> np
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoencoderKLWan, WanImageToVideoPipeline
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> diffusers.utils <span class="hljs-keyword">import</span> export_to_video, load_image
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> CLIPVisionModel

<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-comment"># Available models: Wan-AI/Wan2.1-I2V-14B-480P-Diffusers, Wan-AI/Wan2.1-I2V-14B-720P-Diffusers</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>model_id = <span class="hljs-string">&quot;Wan-AI/Wan2.1-I2V-14B-480P-Diffusers&quot;</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>image_encoder = CLIPVisionModel.from_pretrained(
<span class="hljs-meta">... </span>    model_id, subfolder=<span class="hljs-string">&quot;image_encoder&quot;</span>, torch_dtype=torch.float32
<span class="hljs-meta">... </span>)
<span class="hljs-meta">&gt;&gt;&gt; </span>vae = AutoencoderKLWan.from_pretrained(model_id, subfolder=<span class="hljs-string">&quot;vae&quot;</span>, torch_dtype=torch.float32)
<span class="hljs-meta">&gt;&gt;&gt; </span>pipe = WanImageToVideoPipeline.from_pretrained(
<span class="hljs-meta">... </span>    model_id, vae=vae, image_encoder=image_encoder, torch_dtype=torch.bfloat16
<span class="hljs-meta">... </span>)
<span class="hljs-meta">&gt;&gt;&gt; </span>pipe.to(<span class="hljs-string">&quot;cuda&quot;</span>)

<span class="hljs-meta">&gt;&gt;&gt; </span>image = load_image(
<span class="hljs-meta">... </span>    <span class="hljs-string">&quot;https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/astronaut.jpg&quot;</span>
<span class="hljs-meta">... </span>)
<span class="hljs-meta">&gt;&gt;&gt; </span>max_area = <span class="hljs-number">480</span> * <span class="hljs-number">832</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>aspect_ratio = image.height / image.width
<span class="hljs-meta">&gt;&gt;&gt; </span>mod_value = pipe.vae_scale_factor_spatial * pipe.transformer.config.patch_size[<span class="hljs-number">1</span>]
<span class="hljs-meta">&gt;&gt;&gt; </span>height = <span class="hljs-built_in">round</span>(np.sqrt(max_area * aspect_ratio)) // mod_value * mod_value
<span class="hljs-meta">&gt;&gt;&gt; </span>width = <span class="hljs-built_in">round</span>(np.sqrt(max_area / aspect_ratio)) // mod_value * mod_value
<span class="hljs-meta">&gt;&gt;&gt; </span>image = image.resize((width, height))
<span class="hljs-meta">&gt;&gt;&gt; </span>prompt = (
<span class="hljs-meta">... </span>    <span class="hljs-string">&quot;An astronaut hatching from an egg, on the surface of the moon, the darkness and depth of space realised in &quot;</span>
<span class="hljs-meta">... </span>    <span class="hljs-string">&quot;the background. High quality, ultrarealistic detail and breath-taking movie-like camera shot.&quot;</span>
<span class="hljs-meta">... </span>)
<span class="hljs-meta">&gt;&gt;&gt; </span>negative_prompt = <span class="hljs-string">&quot;Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards&quot;</span>

<span class="hljs-meta">&gt;&gt;&gt; </span>output = pipe(
<span class="hljs-meta">... </span>    image=image,
<span class="hljs-meta">... </span>    prompt=prompt,
<span class="hljs-meta">... </span>    negative_prompt=negative_prompt,
<span class="hljs-meta">... </span>    height=height,
<span class="hljs-meta">... </span>    width=width,
<span class="hljs-meta">... </span>    num_frames=<span class="hljs-number">81</span>,
<span class="hljs-meta">... </span>    guidance_scale=<span class="hljs-number">5.0</span>,
<span class="hljs-meta">... </span>).frames[<span class="hljs-number">0</span>]
<span class="hljs-meta">&gt;&gt;&gt; </span>export_to_video(output, <span class="hljs-string">&quot;output.mp4&quot;</span>, fps=<span class="hljs-number">16</span>)`,wrap:!1}}),{c(){n=p("p"),n.textContent=w,m=i(),y(d.$$.fragment)},l(t){n=c(t,"P",{"data-svelte-h":!0}),f(n)!=="svelte-kvfsh7"&&(n.textContent=w),m=r(t),h(d.$$.fragment,t)},m(t,M){u(t,n,M),u(t,m,M),g(d,t,M),l=!0},p:Xe,i(t){l||(b(d.$$.fragment,t),l=!0)},o(t){J(d.$$.fragment,t),l=!1},d(t){t&&(s(n),s(m)),_(d,t)}}}function ln(W){let n,w,m,d,l,t='<div class="flex flex-wrap space-x-1"><a href="https://huggingface.co/docs/diffusers/main/en/tutorials/using_peft_for_inference" target="_blank" rel="noopener"><img alt="LoRA" src="https://img.shields.io/badge/LoRA-d8b4fe?style=flat"/></a></div>',M,U,T,B,Ut='<a href="https://files.alicdn.com/tpsservice/5c9de1c74de03972b7aa657e5a54756b.pdf" rel="nofollow">Wan2.1</a> is a series of large diffusion transformer available in two versions, a high-performance 14B parameter model and a more accessible 1.3B version. Trained on billions of images and videos, it supports tasks like text-to-video (T2V) and image-to-video (I2V) while enabling features such as camera control and stylistic diversity. The Wan-VAE features better image data compression and a feature cache mechanism that encodes and decodes a video in chunks. To maintain continuity, features from previous chunks are cached and reused for processing subsequent chunks. This improves inference efficiency by reducing memory usage. Wan2.1 also uses a multilingual text encoder and the diffusion transformer models space and time relationships and text conditions with each time step to capture more complex video dynamics.',xe,$,Zt='You can find all the original Wan2.1 checkpoints under the <a href="https://huggingface.co/Wan-AI" rel="nofollow">Wan-AI</a> organization.',Qe,R,Re,z,jt="The example below demonstrates how to generate a video from text optimized for memory or inference speed.",Ye,Y,Ne,S,Fe,v,P,Me,Wt='Wan2.1 supports LoRAs with <a href="/docs/diffusers/pr_11660/en/api/loaders/lora#diffusers.loaders.WanLoraLoaderMixin.load_lora_weights">load_lora_weights()</a>.',tt,A,fe,vt="Show example code",nt,L,st,q,ye,It='<a href="/docs/diffusers/pr_11660/en/api/models/wan_transformer_3d#diffusers.WanTransformer3DModel">WanTransformer3DModel</a> and <a href="/docs/diffusers/pr_11660/en/api/models/autoencoder_kl_wan#diffusers.AutoencoderKLWan">AutoencoderKLWan</a> supports loading from single files with <a href="/docs/diffusers/pr_11660/en/api/loaders/single_file#diffusers.loaders.FromSingleFileMixin.from_single_file">from_single_file()</a>.',at,D,he,Bt="Show example code",ot,K,lt,ge,Gt='<p>Set the <a href="/docs/diffusers/pr_11660/en/api/models/autoencoder_kl_wan#diffusers.AutoencoderKLWan">AutoencoderKLWan</a> dtype to <code>torch.float32</code> for better decoding quality.</p>',it,be,kt="<p>The number of frames per second (fps) or <code>k</code> should be calculated by <code>4 * k + 1</code>.</p>",rt,Je,Vt="<p>Try lower <code>shift</code> values (<code>2.0</code> to <code>5.0</code>) for lower resolution videos and higher <code>shift</code> values (<code>7.0</code> to <code>12.0</code>) for higher resolution images.</p>",He,O,Ee,Z,ee,dt,_e,Ct="Pipeline for text-to-video generation using Wan.",pt,we,Xt=`This model inherits from <a href="/docs/diffusers/pr_11660/en/api/pipelines/overview#diffusers.DiffusionPipeline">DiffusionPipeline</a>. Check the superclass documentation for the generic methods
implemented for all pipelines (downloading, saving, running on a particular device, etc.).`,ct,k,te,mt,Te,xt="The call function to the pipeline for generation.",ut,N,Mt,F,ne,ft,Ue,Qt="Encodes the prompt into text encoder hidden states.",$e,se,ze,j,ae,yt,Ze,Rt="Pipeline for image-to-video generation using Wan.",ht,je,Yt=`This model inherits from <a href="/docs/diffusers/pr_11660/en/api/pipelines/overview#diffusers.DiffusionPipeline">DiffusionPipeline</a>. Check the superclass documentation for the generic methods
implemented for all pipelines (downloading, saving, running on a particular device, etc.).`,gt,V,oe,bt,We,Nt="The call function to the pipeline for generation.",Jt,H,_t,E,le,wt,ve,Ft="Encodes the prompt into text encoder hidden states.",Se,ie,Pe,x,re,Tt,Ie,Ht="Output class for Wan pipelines.",Ae,de,Le,Ve,qe;return U=new Ce({props:{title:"Wan2.1",local:"wan21",headingTag:"h1"}}),R=new Dt({props:{warning:!1,$$slots:{default:[en]},$$scope:{ctx:W}}}),Y=new Ot({props:{id:"usage",options:["memory","inference speed"],$$slots:{default:[sn]},$$scope:{ctx:W}}}),S=new Ce({props:{title:"Notes",local:"notes",headingTag:"h2"}}),L=new ke({props:{code:"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",highlighted:`<span class="hljs-comment"># pip install ftfy</span>
<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoModel, WanPipeline
<span class="hljs-keyword">from</span> diffusers.schedulers.scheduling_unipc_multistep <span class="hljs-keyword">import</span> UniPCMultistepScheduler
<span class="hljs-keyword">from</span> diffusers.utils <span class="hljs-keyword">import</span> export_to_video

vae = AutoModel.from_pretrained(
    <span class="hljs-string">&quot;Wan-AI/Wan2.1-T2V-1.3B-Diffusers&quot;</span>, subfolder=<span class="hljs-string">&quot;vae&quot;</span>, torch_dtype=torch.float32
)
pipeline = WanPipeline.from_pretrained(
    <span class="hljs-string">&quot;Wan-AI/Wan2.1-T2V-1.3B-Diffusers&quot;</span>, vae=vae, torch_dtype=torch.bfloat16
)
pipeline.scheduler = UniPCMultistepScheduler.from_config(
    pipeline.scheduler.config, flow_shift=<span class="hljs-number">5.0</span>
)
pipeline.to(<span class="hljs-string">&quot;cuda&quot;</span>)

pipeline.load_lora_weights(<span class="hljs-string">&quot;benjamin-paine/steamboat-willie-1.3b&quot;</span>, adapter_name=<span class="hljs-string">&quot;steamboat-willie&quot;</span>)
pipeline.set_adapters(<span class="hljs-string">&quot;steamboat-willie&quot;</span>)

pipeline.enable_model_cpu_offload()

<span class="hljs-comment"># use &quot;steamboat willie style&quot; to trigger the LoRA</span>
prompt = <span class="hljs-string">&quot;&quot;&quot;
steamboat willie style, golden era animation, The camera rushes from far to near in a low-angle shot, 
revealing a white ferret on a log. It plays, leaps into the water, and emerges, as the camera zooms in 
for a close-up. Water splashes berry bushes nearby, while moss, snow, and leaves blanket the ground. 
Birch trees and a light blue sky frame the scene, with ferns in the foreground. Side lighting casts dynamic 
shadows and warm highlights. Medium composition, front view, low angle, with depth of field.
&quot;&quot;&quot;</span>

output = pipeline(
    prompt=prompt,
    num_frames=<span class="hljs-number">81</span>,
    guidance_scale=<span class="hljs-number">5.0</span>,
).frames[<span class="hljs-number">0</span>]
export_to_video(output, <span class="hljs-string">&quot;output.mp4&quot;</span>, fps=<span class="hljs-number">16</span>)`,wrap:!1}}),K=new ke({props:{code:"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",highlighted:`<span class="hljs-comment"># pip install ftfy</span>
<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> WanPipeline, AutoModel

vae = AutoModel.from_single_file(
    <span class="hljs-string">&quot;https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors&quot;</span>
)
transformer = AutoModel.from_single_file(
    <span class="hljs-string">&quot;https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_t2v_1.3B_bf16.safetensors&quot;</span>,
    torch_dtype=torch.bfloat16
)
pipeline = WanPipeline.from_pretrained(
    <span class="hljs-string">&quot;Wan-AI/Wan2.1-T2V-1.3B-Diffusers&quot;</span>,
    vae=vae,
    transformer=transformer,
    torch_dtype=torch.bfloat16
)`,wrap:!1}}),O=new Ce({props:{title:"WanPipeline",local:"diffusers.WanPipeline",headingTag:"h2"}}),ee=new ue({props:{name:"class diffusers.WanPipeline",anchor:"diffusers.WanPipeline",parameters:[{name:"tokenizer",val:": AutoTokenizer"},{name:"text_encoder",val:": UMT5EncoderModel"},{name:"transformer",val:": WanTransformer3DModel"},{name:"vae",val:": AutoencoderKLWan"},{name:"scheduler",val:": FlowMatchEulerDiscreteScheduler"}],parametersDescription:[{anchor:"diffusers.WanPipeline.tokenizer",description:`<strong>tokenizer</strong> (<code>T5Tokenizer</code>) &#x2014;
Tokenizer from <a href="https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5Tokenizer" rel="nofollow">T5</a>,
specifically the <a href="https://huggingface.co/google/umt5-xxl" rel="nofollow">google/umt5-xxl</a> variant.`,name:"tokenizer"},{anchor:"diffusers.WanPipeline.text_encoder",description:`<strong>text_encoder</strong> (<code>T5EncoderModel</code>) &#x2014;
<a href="https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5EncoderModel" rel="nofollow">T5</a>, specifically
the <a href="https://huggingface.co/google/umt5-xxl" rel="nofollow">google/umt5-xxl</a> variant.`,name:"text_encoder"},{anchor:"diffusers.WanPipeline.transformer",description:`<strong>transformer</strong> (<a href="/docs/diffusers/pr_11660/en/api/models/wan_transformer_3d#diffusers.WanTransformer3DModel">WanTransformer3DModel</a>) &#x2014;
Conditional Transformer to denoise the input latents.`,name:"transformer"},{anchor:"diffusers.WanPipeline.scheduler",description:`<strong>scheduler</strong> (<a href="/docs/diffusers/pr_11660/en/api/schedulers/unipc#diffusers.UniPCMultistepScheduler">UniPCMultistepScheduler</a>) &#x2014;
A scheduler to be used in combination with <code>transformer</code> to denoise the encoded image latents.`,name:"scheduler"},{anchor:"diffusers.WanPipeline.vae",description:`<strong>vae</strong> (<a href="/docs/diffusers/pr_11660/en/api/models/autoencoder_kl_wan#diffusers.AutoencoderKLWan">AutoencoderKLWan</a>) &#x2014;
Variational Auto-Encoder (VAE) Model to encode and decode videos to and from latent representations.`,name:"vae"}],source:"https://github.com/huggingface/diffusers/blob/vr_11660/src/diffusers/pipelines/wan/pipeline_wan.py#L95"}}),te=new ue({props:{name:"__call__",anchor:"diffusers.WanPipeline.__call__",parameters:[{name:"prompt",val:": typing.Union[str, typing.List[str]] = None"},{name:"negative_prompt",val:": typing.Union[str, typing.List[str]] = None"},{name:"height",val:": int = 480"},{name:"width",val:": int = 832"},{name:"num_frames",val:": int = 81"},{name:"num_inference_steps",val:": int = 50"},{name:"guidance_scale",val:": float = 5.0"},{name:"num_videos_per_prompt",val:": typing.Optional[int] = 1"},{name:"generator",val:": typing.Union[torch._C.Generator, typing.List[torch._C.Generator], NoneType] = None"},{name:"latents",val:": typing.Optional[torch.Tensor] = None"},{name:"prompt_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"negative_prompt_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"output_type",val:": typing.Optional[str] = 'np'"},{name:"return_dict",val:": bool = True"},{name:"attention_kwargs",val:": typing.Optional[typing.Dict[str, typing.Any]] = None"},{name:"callback_on_step_end",val:": typing.Union[typing.Callable[[int, int, typing.Dict], NoneType], diffusers.callbacks.PipelineCallback, diffusers.callbacks.MultiPipelineCallbacks, NoneType] = None"},{name:"callback_on_step_end_tensor_inputs",val:": typing.List[str] = ['latents']"},{name:"max_sequence_length",val:": int = 512"}],parametersDescription:[{anchor:"diffusers.WanPipeline.__call__.prompt",description:`<strong>prompt</strong> (<code>str</code> or <code>List[str]</code>, <em>optional</em>) &#x2014;
The prompt or prompts to guide the image generation. If not defined, one has to pass <code>prompt_embeds</code>.
instead.`,name:"prompt"},{anchor:"diffusers.WanPipeline.__call__.height",description:`<strong>height</strong> (<code>int</code>, defaults to <code>480</code>) &#x2014;
The height in pixels of the generated image.`,name:"height"},{anchor:"diffusers.WanPipeline.__call__.width",description:`<strong>width</strong> (<code>int</code>, defaults to <code>832</code>) &#x2014;
The width in pixels of the generated image.`,name:"width"},{anchor:"diffusers.WanPipeline.__call__.num_frames",description:`<strong>num_frames</strong> (<code>int</code>, defaults to <code>81</code>) &#x2014;
The number of frames in the generated video.`,name:"num_frames"},{anchor:"diffusers.WanPipeline.__call__.num_inference_steps",description:`<strong>num_inference_steps</strong> (<code>int</code>, defaults to <code>50</code>) &#x2014;
The number of denoising steps. More denoising steps usually lead to a higher quality image at the
expense of slower inference.`,name:"num_inference_steps"},{anchor:"diffusers.WanPipeline.__call__.guidance_scale",description:`<strong>guidance_scale</strong> (<code>float</code>, defaults to <code>5.0</code>) &#x2014;
Guidance scale as defined in <a href="https://huggingface.co/papers/2207.12598" rel="nofollow">Classifier-Free Diffusion
Guidance</a>. <code>guidance_scale</code> is defined as <code>w</code> of equation 2.
of <a href="https://huggingface.co/papers/2205.11487" rel="nofollow">Imagen Paper</a>. Guidance scale is enabled by setting
<code>guidance_scale &gt; 1</code>. Higher guidance scale encourages to generate images that are closely linked to
the text <code>prompt</code>, usually at the expense of lower image quality.`,name:"guidance_scale"},{anchor:"diffusers.WanPipeline.__call__.num_videos_per_prompt",description:`<strong>num_videos_per_prompt</strong> (<code>int</code>, <em>optional</em>, defaults to 1) &#x2014;
The number of images to generate per prompt.`,name:"num_videos_per_prompt"},{anchor:"diffusers.WanPipeline.__call__.generator",description:`<strong>generator</strong> (<code>torch.Generator</code> or <code>List[torch.Generator]</code>, <em>optional</em>) &#x2014;
A <a href="https://pytorch.org/docs/stable/generated/torch.Generator.html" rel="nofollow"><code>torch.Generator</code></a> to make
generation deterministic.`,name:"generator"},{anchor:"diffusers.WanPipeline.__call__.latents",description:`<strong>latents</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for image
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 <code>generator</code>.`,name:"latents"},{anchor:"diffusers.WanPipeline.__call__.prompt_embeds",description:`<strong>prompt_embeds</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated text embeddings. Can be used to easily tweak text inputs (prompt weighting). If not
provided, text embeddings are generated from the <code>prompt</code> input argument.`,name:"prompt_embeds"},{anchor:"diffusers.WanPipeline.__call__.output_type",description:`<strong>output_type</strong> (<code>str</code>, <em>optional</em>, defaults to <code>&quot;np&quot;</code>) &#x2014;
The output format of the generated image. Choose between <code>PIL.Image</code> or <code>np.array</code>.`,name:"output_type"},{anchor:"diffusers.WanPipeline.__call__.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) &#x2014;
Whether or not to return a <code>WanPipelineOutput</code> instead of a plain tuple.`,name:"return_dict"},{anchor:"diffusers.WanPipeline.__call__.attention_kwargs",description:`<strong>attention_kwargs</strong> (<code>dict</code>, <em>optional</em>) &#x2014;
A kwargs dictionary that if specified is passed along to the <code>AttentionProcessor</code> as defined under
<code>self.processor</code> in
<a href="https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py" rel="nofollow">diffusers.models.attention_processor</a>.`,name:"attention_kwargs"},{anchor:"diffusers.WanPipeline.__call__.callback_on_step_end",description:`<strong>callback_on_step_end</strong> (<code>Callable</code>, <code>PipelineCallback</code>, <code>MultiPipelineCallbacks</code>, <em>optional</em>) &#x2014;
A function or a subclass of <code>PipelineCallback</code> or <code>MultiPipelineCallbacks</code> that is called at the end of
each denoising step during the inference. with the following arguments: <code>callback_on_step_end(self: DiffusionPipeline, step: int, timestep: int, callback_kwargs: Dict)</code>. <code>callback_kwargs</code> will include a
list of all tensors as specified by <code>callback_on_step_end_tensor_inputs</code>.`,name:"callback_on_step_end"},{anchor:"diffusers.WanPipeline.__call__.callback_on_step_end_tensor_inputs",description:`<strong>callback_on_step_end_tensor_inputs</strong> (<code>List</code>, <em>optional</em>) &#x2014;
The list of tensor inputs for the <code>callback_on_step_end</code> function. The tensors specified in the list
will be passed as <code>callback_kwargs</code> argument. You will only be able to include variables listed in the
<code>._callback_tensor_inputs</code> attribute of your pipeline class.`,name:"callback_on_step_end_tensor_inputs"},{anchor:"diffusers.WanPipeline.__call__.autocast_dtype",description:`<strong>autocast_dtype</strong> (<code>torch.dtype</code>, <em>optional</em>, defaults to <code>torch.bfloat16</code>) &#x2014;
The dtype to use for the torch.amp.autocast.`,name:"autocast_dtype"}],source:"https://github.com/huggingface/diffusers/blob/vr_11660/src/diffusers/pipelines/wan/pipeline_wan.py#L361",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


<p>If <code>return_dict</code> is <code>True</code>, <code>WanPipelineOutput</code> is returned, otherwise a <code>tuple</code> is returned where
the first element is a list with the generated images and the second element is a list of <code>bool</code>s
indicating whether the corresponding generated image contains “not-safe-for-work” (nsfw) content.</p>
`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


<p><code>~WanPipelineOutput</code> or <code>tuple</code></p>
`}}),N=new Et({props:{anchor:"diffusers.WanPipeline.__call__.example",$$slots:{default:[an]},$$scope:{ctx:W}}}),ne=new ue({props:{name:"encode_prompt",anchor:"diffusers.WanPipeline.encode_prompt",parameters:[{name:"prompt",val:": typing.Union[str, typing.List[str]]"},{name:"negative_prompt",val:": typing.Union[str, typing.List[str], NoneType] = None"},{name:"do_classifier_free_guidance",val:": bool = True"},{name:"num_videos_per_prompt",val:": int = 1"},{name:"prompt_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"negative_prompt_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"max_sequence_length",val:": int = 226"},{name:"device",val:": typing.Optional[torch.device] = None"},{name:"dtype",val:": typing.Optional[torch.dtype] = None"}],parametersDescription:[{anchor:"diffusers.WanPipeline.encode_prompt.prompt",description:`<strong>prompt</strong> (<code>str</code> or <code>List[str]</code>, <em>optional</em>) &#x2014;
prompt to be encoded`,name:"prompt"},{anchor:"diffusers.WanPipeline.encode_prompt.negative_prompt",description:`<strong>negative_prompt</strong> (<code>str</code> or <code>List[str]</code>, <em>optional</em>) &#x2014;
The prompt or prompts not to guide the image generation. If not defined, one has to pass
<code>negative_prompt_embeds</code> instead. Ignored when not using guidance (i.e., ignored if <code>guidance_scale</code> is
less than <code>1</code>).`,name:"negative_prompt"},{anchor:"diffusers.WanPipeline.encode_prompt.do_classifier_free_guidance",description:`<strong>do_classifier_free_guidance</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) &#x2014;
Whether to use classifier free guidance or not.`,name:"do_classifier_free_guidance"},{anchor:"diffusers.WanPipeline.encode_prompt.num_videos_per_prompt",description:`<strong>num_videos_per_prompt</strong> (<code>int</code>, <em>optional</em>, defaults to 1) &#x2014;
Number of videos that should be generated per prompt. torch device to place the resulting embeddings on`,name:"num_videos_per_prompt"},{anchor:"diffusers.WanPipeline.encode_prompt.prompt_embeds",description:`<strong>prompt_embeds</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated text embeddings. Can be used to easily tweak text inputs, <em>e.g.</em> prompt weighting. If not
provided, text embeddings will be generated from <code>prompt</code> input argument.`,name:"prompt_embeds"},{anchor:"diffusers.WanPipeline.encode_prompt.negative_prompt_embeds",description:`<strong>negative_prompt_embeds</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated negative text embeddings. Can be used to easily tweak text inputs, <em>e.g.</em> prompt
weighting. If not provided, negative_prompt_embeds will be generated from <code>negative_prompt</code> input
argument.`,name:"negative_prompt_embeds"},{anchor:"diffusers.WanPipeline.encode_prompt.device",description:`<strong>device</strong> &#x2014; (<code>torch.device</code>, <em>optional</em>):
torch device`,name:"device"},{anchor:"diffusers.WanPipeline.encode_prompt.dtype",description:`<strong>dtype</strong> &#x2014; (<code>torch.dtype</code>, <em>optional</em>):
torch dtype`,name:"dtype"}],source:"https://github.com/huggingface/diffusers/blob/vr_11660/src/diffusers/pipelines/wan/pipeline_wan.py#L183"}}),se=new Ce({props:{title:"WanImageToVideoPipeline",local:"diffusers.WanImageToVideoPipeline",headingTag:"h2"}}),ae=new ue({props:{name:"class diffusers.WanImageToVideoPipeline",anchor:"diffusers.WanImageToVideoPipeline",parameters:[{name:"tokenizer",val:": AutoTokenizer"},{name:"text_encoder",val:": UMT5EncoderModel"},{name:"image_encoder",val:": CLIPVisionModel"},{name:"image_processor",val:": CLIPImageProcessor"},{name:"transformer",val:": WanTransformer3DModel"},{name:"vae",val:": AutoencoderKLWan"},{name:"scheduler",val:": FlowMatchEulerDiscreteScheduler"}],parametersDescription:[{anchor:"diffusers.WanImageToVideoPipeline.tokenizer",description:`<strong>tokenizer</strong> (<code>T5Tokenizer</code>) &#x2014;
Tokenizer from <a href="https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5Tokenizer" rel="nofollow">T5</a>,
specifically the <a href="https://huggingface.co/google/umt5-xxl" rel="nofollow">google/umt5-xxl</a> variant.`,name:"tokenizer"},{anchor:"diffusers.WanImageToVideoPipeline.text_encoder",description:`<strong>text_encoder</strong> (<code>T5EncoderModel</code>) &#x2014;
<a href="https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5EncoderModel" rel="nofollow">T5</a>, specifically
the <a href="https://huggingface.co/google/umt5-xxl" rel="nofollow">google/umt5-xxl</a> variant.`,name:"text_encoder"},{anchor:"diffusers.WanImageToVideoPipeline.image_encoder",description:`<strong>image_encoder</strong> (<code>CLIPVisionModel</code>) &#x2014;
<a href="https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPVisionModel" rel="nofollow">CLIP</a>, specifically
the
<a href="https://github.com/mlfoundations/open_clip/blob/main/docs/PRETRAINED.md#vit-h14-xlm-roberta-large" rel="nofollow">clip-vit-huge-patch14</a>
variant.`,name:"image_encoder"},{anchor:"diffusers.WanImageToVideoPipeline.transformer",description:`<strong>transformer</strong> (<a href="/docs/diffusers/pr_11660/en/api/models/wan_transformer_3d#diffusers.WanTransformer3DModel">WanTransformer3DModel</a>) &#x2014;
Conditional Transformer to denoise the input latents.`,name:"transformer"},{anchor:"diffusers.WanImageToVideoPipeline.scheduler",description:`<strong>scheduler</strong> (<a href="/docs/diffusers/pr_11660/en/api/schedulers/unipc#diffusers.UniPCMultistepScheduler">UniPCMultistepScheduler</a>) &#x2014;
A scheduler to be used in combination with <code>transformer</code> to denoise the encoded image latents.`,name:"scheduler"},{anchor:"diffusers.WanImageToVideoPipeline.vae",description:`<strong>vae</strong> (<a href="/docs/diffusers/pr_11660/en/api/models/autoencoder_kl_wan#diffusers.AutoencoderKLWan">AutoencoderKLWan</a>) &#x2014;
Variational Auto-Encoder (VAE) Model to encode and decode videos to and from latent representations.`,name:"vae"}],source:"https://github.com/huggingface/diffusers/blob/vr_11660/src/diffusers/pipelines/wan/pipeline_wan_i2v.py#L127"}}),oe=new ue({props:{name:"__call__",anchor:"diffusers.WanImageToVideoPipeline.__call__",parameters:[{name:"image",val:": typing.Union[PIL.Image.Image, numpy.ndarray, torch.Tensor, typing.List[PIL.Image.Image], typing.List[numpy.ndarray], typing.List[torch.Tensor]]"},{name:"prompt",val:": typing.Union[str, typing.List[str]] = None"},{name:"negative_prompt",val:": typing.Union[str, typing.List[str]] = None"},{name:"height",val:": int = 480"},{name:"width",val:": int = 832"},{name:"num_frames",val:": int = 81"},{name:"num_inference_steps",val:": int = 50"},{name:"guidance_scale",val:": float = 5.0"},{name:"num_videos_per_prompt",val:": typing.Optional[int] = 1"},{name:"generator",val:": typing.Union[torch._C.Generator, typing.List[torch._C.Generator], NoneType] = None"},{name:"latents",val:": typing.Optional[torch.Tensor] = None"},{name:"prompt_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"negative_prompt_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"image_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"last_image",val:": typing.Optional[torch.Tensor] = None"},{name:"output_type",val:": typing.Optional[str] = 'np'"},{name:"return_dict",val:": bool = True"},{name:"attention_kwargs",val:": typing.Optional[typing.Dict[str, typing.Any]] = None"},{name:"callback_on_step_end",val:": typing.Union[typing.Callable[[int, int, typing.Dict], NoneType], diffusers.callbacks.PipelineCallback, diffusers.callbacks.MultiPipelineCallbacks, NoneType] = None"},{name:"callback_on_step_end_tensor_inputs",val:": typing.List[str] = ['latents']"},{name:"max_sequence_length",val:": int = 512"}],parametersDescription:[{anchor:"diffusers.WanImageToVideoPipeline.__call__.image",description:`<strong>image</strong> (<code>PipelineImageInput</code>) &#x2014;
The input image to condition the generation on. Must be an image, a list of images or a <code>torch.Tensor</code>.`,name:"image"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.prompt",description:`<strong>prompt</strong> (<code>str</code> or <code>List[str]</code>, <em>optional</em>) &#x2014;
The prompt or prompts to guide the image generation. If not defined, one has to pass <code>prompt_embeds</code>.
instead.`,name:"prompt"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.negative_prompt",description:`<strong>negative_prompt</strong> (<code>str</code> or <code>List[str]</code>, <em>optional</em>) &#x2014;
The prompt or prompts not to guide the image generation. If not defined, one has to pass
<code>negative_prompt_embeds</code> instead. Ignored when not using guidance (i.e., ignored if <code>guidance_scale</code> is
less than <code>1</code>).`,name:"negative_prompt"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.height",description:`<strong>height</strong> (<code>int</code>, defaults to <code>480</code>) &#x2014;
The height of the generated video.`,name:"height"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.width",description:`<strong>width</strong> (<code>int</code>, defaults to <code>832</code>) &#x2014;
The width of the generated video.`,name:"width"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.num_frames",description:`<strong>num_frames</strong> (<code>int</code>, defaults to <code>81</code>) &#x2014;
The number of frames in the generated video.`,name:"num_frames"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.num_inference_steps",description:`<strong>num_inference_steps</strong> (<code>int</code>, defaults to <code>50</code>) &#x2014;
The number of denoising steps. More denoising steps usually lead to a higher quality image at the
expense of slower inference.`,name:"num_inference_steps"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.guidance_scale",description:`<strong>guidance_scale</strong> (<code>float</code>, defaults to <code>5.0</code>) &#x2014;
Guidance scale as defined in <a href="https://huggingface.co/papers/2207.12598" rel="nofollow">Classifier-Free Diffusion
Guidance</a>. <code>guidance_scale</code> is defined as <code>w</code> of equation 2.
of <a href="https://huggingface.co/papers/2205.11487" rel="nofollow">Imagen Paper</a>. Guidance scale is enabled by setting
<code>guidance_scale &gt; 1</code>. Higher guidance scale encourages to generate images that are closely linked to
the text <code>prompt</code>, usually at the expense of lower image quality.`,name:"guidance_scale"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.num_videos_per_prompt",description:`<strong>num_videos_per_prompt</strong> (<code>int</code>, <em>optional</em>, defaults to 1) &#x2014;
The number of images to generate per prompt.`,name:"num_videos_per_prompt"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.generator",description:`<strong>generator</strong> (<code>torch.Generator</code> or <code>List[torch.Generator]</code>, <em>optional</em>) &#x2014;
A <a href="https://pytorch.org/docs/stable/generated/torch.Generator.html" rel="nofollow"><code>torch.Generator</code></a> to make
generation deterministic.`,name:"generator"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.latents",description:`<strong>latents</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for image
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 <code>generator</code>.`,name:"latents"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.prompt_embeds",description:`<strong>prompt_embeds</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated text embeddings. Can be used to easily tweak text inputs (prompt weighting). If not
provided, text embeddings are generated from the <code>prompt</code> input argument.`,name:"prompt_embeds"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.negative_prompt_embeds",description:`<strong>negative_prompt_embeds</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated text embeddings. Can be used to easily tweak text inputs (prompt weighting). If not
provided, text embeddings are generated from the <code>negative_prompt</code> input argument.`,name:"negative_prompt_embeds"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.image_embeds",description:`<strong>image_embeds</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated image embeddings. Can be used to easily tweak image inputs (weighting). If not provided,
image embeddings are generated from the <code>image</code> input argument.`,name:"image_embeds"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.output_type",description:`<strong>output_type</strong> (<code>str</code>, <em>optional</em>, defaults to <code>&quot;np&quot;</code>) &#x2014;
The output format of the generated image. Choose between <code>PIL.Image</code> or <code>np.array</code>.`,name:"output_type"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) &#x2014;
Whether or not to return a <code>WanPipelineOutput</code> instead of a plain tuple.`,name:"return_dict"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.attention_kwargs",description:`<strong>attention_kwargs</strong> (<code>dict</code>, <em>optional</em>) &#x2014;
A kwargs dictionary that if specified is passed along to the <code>AttentionProcessor</code> as defined under
<code>self.processor</code> in
<a href="https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py" rel="nofollow">diffusers.models.attention_processor</a>.`,name:"attention_kwargs"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.callback_on_step_end",description:`<strong>callback_on_step_end</strong> (<code>Callable</code>, <code>PipelineCallback</code>, <code>MultiPipelineCallbacks</code>, <em>optional</em>) &#x2014;
A function or a subclass of <code>PipelineCallback</code> or <code>MultiPipelineCallbacks</code> that is called at the end of
each denoising step during the inference. with the following arguments: <code>callback_on_step_end(self: DiffusionPipeline, step: int, timestep: int, callback_kwargs: Dict)</code>. <code>callback_kwargs</code> will include a
list of all tensors as specified by <code>callback_on_step_end_tensor_inputs</code>.`,name:"callback_on_step_end"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.callback_on_step_end_tensor_inputs",description:`<strong>callback_on_step_end_tensor_inputs</strong> (<code>List</code>, <em>optional</em>) &#x2014;
The list of tensor inputs for the <code>callback_on_step_end</code> function. The tensors specified in the list
will be passed as <code>callback_kwargs</code> argument. You will only be able to include variables listed in the
<code>._callback_tensor_inputs</code> attribute of your pipeline class.`,name:"callback_on_step_end_tensor_inputs"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.max_sequence_length",description:`<strong>max_sequence_length</strong> (<code>int</code>, <em>optional</em>, defaults to <code>512</code>) &#x2014;
The maximum sequence length of the prompt.`,name:"max_sequence_length"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.shift",description:`<strong>shift</strong> (<code>float</code>, <em>optional</em>, defaults to <code>5.0</code>) &#x2014;
The shift of the flow.`,name:"shift"},{anchor:"diffusers.WanImageToVideoPipeline.__call__.autocast_dtype",description:`<strong>autocast_dtype</strong> (<code>torch.dtype</code>, <em>optional</em>, defaults to <code>torch.bfloat16</code>) &#x2014;
The dtype to use for the torch.amp.autocast.`,name:"autocast_dtype"}],source:"https://github.com/huggingface/diffusers/blob/vr_11660/src/diffusers/pipelines/wan/pipeline_wan_i2v.py#L474",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


<p>If <code>return_dict</code> is <code>True</code>, <code>WanPipelineOutput</code> is returned, otherwise a <code>tuple</code> is returned where
the first element is a list with the generated images and the second element is a list of <code>bool</code>s
indicating whether the corresponding generated image contains “not-safe-for-work” (nsfw) content.</p>
`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


<p><code>~WanPipelineOutput</code> or <code>tuple</code></p>
`}}),H=new Et({props:{anchor:"diffusers.WanImageToVideoPipeline.__call__.example",$$slots:{default:[on]},$$scope:{ctx:W}}}),le=new ue({props:{name:"encode_prompt",anchor:"diffusers.WanImageToVideoPipeline.encode_prompt",parameters:[{name:"prompt",val:": typing.Union[str, typing.List[str]]"},{name:"negative_prompt",val:": typing.Union[str, typing.List[str], NoneType] = None"},{name:"do_classifier_free_guidance",val:": bool = True"},{name:"num_videos_per_prompt",val:": int = 1"},{name:"prompt_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"negative_prompt_embeds",val:": typing.Optional[torch.Tensor] = None"},{name:"max_sequence_length",val:": int = 226"},{name:"device",val:": typing.Optional[torch.device] = None"},{name:"dtype",val:": typing.Optional[torch.dtype] = None"}],parametersDescription:[{anchor:"diffusers.WanImageToVideoPipeline.encode_prompt.prompt",description:`<strong>prompt</strong> (<code>str</code> or <code>List[str]</code>, <em>optional</em>) &#x2014;
prompt to be encoded`,name:"prompt"},{anchor:"diffusers.WanImageToVideoPipeline.encode_prompt.negative_prompt",description:`<strong>negative_prompt</strong> (<code>str</code> or <code>List[str]</code>, <em>optional</em>) &#x2014;
The prompt or prompts not to guide the image generation. If not defined, one has to pass
<code>negative_prompt_embeds</code> instead. Ignored when not using guidance (i.e., ignored if <code>guidance_scale</code> is
less than <code>1</code>).`,name:"negative_prompt"},{anchor:"diffusers.WanImageToVideoPipeline.encode_prompt.do_classifier_free_guidance",description:`<strong>do_classifier_free_guidance</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) &#x2014;
Whether to use classifier free guidance or not.`,name:"do_classifier_free_guidance"},{anchor:"diffusers.WanImageToVideoPipeline.encode_prompt.num_videos_per_prompt",description:`<strong>num_videos_per_prompt</strong> (<code>int</code>, <em>optional</em>, defaults to 1) &#x2014;
Number of videos that should be generated per prompt. torch device to place the resulting embeddings on`,name:"num_videos_per_prompt"},{anchor:"diffusers.WanImageToVideoPipeline.encode_prompt.prompt_embeds",description:`<strong>prompt_embeds</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated text embeddings. Can be used to easily tweak text inputs, <em>e.g.</em> prompt weighting. If not
provided, text embeddings will be generated from <code>prompt</code> input argument.`,name:"prompt_embeds"},{anchor:"diffusers.WanImageToVideoPipeline.encode_prompt.negative_prompt_embeds",description:`<strong>negative_prompt_embeds</strong> (<code>torch.Tensor</code>, <em>optional</em>) &#x2014;
Pre-generated negative text embeddings. Can be used to easily tweak text inputs, <em>e.g.</em> prompt
weighting. If not provided, negative_prompt_embeds will be generated from <code>negative_prompt</code> input
argument.`,name:"negative_prompt_embeds"},{anchor:"diffusers.WanImageToVideoPipeline.encode_prompt.device",description:`<strong>device</strong> &#x2014; (<code>torch.device</code>, <em>optional</em>):
torch device`,name:"device"},{anchor:"diffusers.WanImageToVideoPipeline.encode_prompt.dtype",description:`<strong>dtype</strong> &#x2014; (<code>torch.dtype</code>, <em>optional</em>):
torch dtype`,name:"dtype"}],source:"https://github.com/huggingface/diffusers/blob/vr_11660/src/diffusers/pipelines/wan/pipeline_wan_i2v.py#L236"}}),ie=new Ce({props:{title:"WanPipelineOutput",local:"diffusers.pipelines.wan.pipeline_output.WanPipelineOutput",headingTag:"h2"}}),re=new ue({props:{name:"class diffusers.pipelines.wan.pipeline_output.WanPipelineOutput",anchor:"diffusers.pipelines.wan.pipeline_output.WanPipelineOutput",parameters:[{name:"frames",val:": Tensor"}],parametersDescription:[{anchor:"diffusers.pipelines.wan.pipeline_output.WanPipelineOutput.frames",description:`<strong>frames</strong> (<code>torch.Tensor</code>, <code>np.ndarray</code>, or List[List[PIL.Image.Image]]) &#x2014;
List of video outputs - It can be a nested list of length <code>batch_size,</code> with each sub-list containing
denoised PIL image sequences of length <code>num_frames.</code> It can also be a NumPy array or Torch tensor of shape
<code>(batch_size, num_frames, channels, height, width)</code>.`,name:"frames"}],source:"https://github.com/huggingface/diffusers/blob/vr_11660/src/diffusers/pipelines/wan/pipeline_output.py#L8"}}),de=new Kt({props:{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/pipelines/wan.md"}}),{c(){n=p("meta"),w=i(),m=p("p"),d=i(),l=p("div"),l.innerHTML=t,M=i(),y(U.$$.fragment),T=i(),B=p("p"),B.innerHTML=Ut,xe=i(),$=p("p"),$.innerHTML=Zt,Qe=i(),y(R.$$.fragment),Re=i(),z=p("p"),z.textContent=jt,Ye=i(),y(Y.$$.fragment),Ne=i(),y(S.$$.fragment),Fe=i(),v=p("ul"),P=p("li"),Me=p("p"),Me.innerHTML=Wt,tt=i(),A=p("details"),fe=p("summary"),fe.textContent=vt,nt=i(),y(L.$$.fragment),st=i(),q=p("li"),ye=p("p"),ye.innerHTML=It,at=i(),D=p("details"),he=p("summary"),he.textContent=Bt,ot=i(),y(K.$$.fragment),lt=i(),ge=p("li"),ge.innerHTML=Gt,it=i(),be=p("li"),be.innerHTML=kt,rt=i(),Je=p("li"),Je.innerHTML=Vt,He=i(),y(O.$$.fragment),Ee=i(),Z=p("div"),y(ee.$$.fragment),dt=i(),_e=p("p"),_e.textContent=Ct,pt=i(),we=p("p"),we.innerHTML=Xt,ct=i(),k=p("div"),y(te.$$.fragment),mt=i(),Te=p("p"),Te.textContent=xt,ut=i(),y(N.$$.fragment),Mt=i(),F=p("div"),y(ne.$$.fragment),ft=i(),Ue=p("p"),Ue.textContent=Qt,$e=i(),y(se.$$.fragment),ze=i(),j=p("div"),y(ae.$$.fragment),yt=i(),Ze=p("p"),Ze.textContent=Rt,ht=i(),je=p("p"),je.innerHTML=Yt,gt=i(),V=p("div"),y(oe.$$.fragment),bt=i(),We=p("p"),We.textContent=Nt,Jt=i(),y(H.$$.fragment),_t=i(),E=p("div"),y(le.$$.fragment),wt=i(),ve=p("p"),ve.textContent=Ft,Se=i(),y(ie.$$.fragment),Pe=i(),x=p("div"),y(re.$$.fragment),Tt=i(),Ie=p("p"),Ie.textContent=Ht,Ae=i(),y(de.$$.fragment),Le=i(),Ve=p("p"),this.h()},l(e){const a=Lt("svelte-u9bgzb",document.head);n=c(a,"META",{name:!0,content:!0}),a.forEach(s),w=r(e),m=c(e,"P",{}),I(m).forEach(s),d=r(e),l=c(e,"DIV",{style:!0,"data-svelte-h":!0}),f(l)!=="svelte-p206qu"&&(l.innerHTML=t),M=r(e),h(U.$$.fragment,e),T=r(e),B=c(e,"P",{"data-svelte-h":!0}),f(B)!=="svelte-1tw7m3e"&&(B.innerHTML=Ut),xe=r(e),$=c(e,"P",{"data-svelte-h":!0}),f($)!=="svelte-1uk9cie"&&($.innerHTML=Zt),Qe=r(e),h(R.$$.fragment,e),Re=r(e),z=c(e,"P",{"data-svelte-h":!0}),f(z)!=="svelte-r8z9s6"&&(z.textContent=jt),Ye=r(e),h(Y.$$.fragment,e),Ne=r(e),h(S.$$.fragment,e),Fe=r(e),v=c(e,"UL",{});var G=I(v);P=c(G,"LI",{});var pe=I(P);Me=c(pe,"P",{"data-svelte-h":!0}),f(Me)!=="svelte-14j6aw9"&&(Me.innerHTML=Wt),tt=r(pe),A=c(pe,"DETAILS",{});var ce=I(A);fe=c(ce,"SUMMARY",{"data-svelte-h":!0}),f(fe)!=="svelte-1m0l1gk"&&(fe.textContent=vt),nt=r(ce),h(L.$$.fragment,ce),ce.forEach(s),pe.forEach(s),st=r(G),q=c(G,"LI",{});var me=I(q);ye=c(me,"P",{"data-svelte-h":!0}),f(ye)!=="svelte-1i1om3x"&&(ye.innerHTML=It),at=r(me),D=c(me,"DETAILS",{});var De=I(D);he=c(De,"SUMMARY",{"data-svelte-h":!0}),f(he)!=="svelte-1m0l1gk"&&(he.textContent=Bt),ot=r(De),h(K.$$.fragment,De),De.forEach(s),me.forEach(s),lt=r(G),ge=c(G,"LI",{"data-svelte-h":!0}),f(ge)!=="svelte-1w7r5pk"&&(ge.innerHTML=Gt),it=r(G),be=c(G,"LI",{"data-svelte-h":!0}),f(be)!=="svelte-154noqv"&&(be.innerHTML=kt),rt=r(G),Je=c(G,"LI",{"data-svelte-h":!0}),f(Je)!=="svelte-ck8fz4"&&(Je.innerHTML=Vt),G.forEach(s),He=r(e),h(O.$$.fragment,e),Ee=r(e),Z=c(e,"DIV",{class:!0});var C=I(Z);h(ee.$$.fragment,C),dt=r(C),_e=c(C,"P",{"data-svelte-h":!0}),f(_e)!=="svelte-cppop6"&&(_e.textContent=Ct),pt=r(C),we=c(C,"P",{"data-svelte-h":!0}),f(we)!=="svelte-1ffteuy"&&(we.innerHTML=Xt),ct=r(C),k=c(C,"DIV",{class:!0});var Be=I(k);h(te.$$.fragment,Be),mt=r(Be),Te=c(Be,"P",{"data-svelte-h":!0}),f(Te)!=="svelte-50j04k"&&(Te.textContent=xt),ut=r(Be),h(N.$$.fragment,Be),Be.forEach(s),Mt=r(C),F=c(C,"DIV",{class:!0});var Ke=I(F);h(ne.$$.fragment,Ke),ft=r(Ke),Ue=c(Ke,"P",{"data-svelte-h":!0}),f(Ue)!=="svelte-16q0ax1"&&(Ue.textContent=Qt),Ke.forEach(s),C.forEach(s),$e=r(e),h(se.$$.fragment,e),ze=r(e),j=c(e,"DIV",{class:!0});var X=I(j);h(ae.$$.fragment,X),yt=r(X),Ze=c(X,"P",{"data-svelte-h":!0}),f(Ze)!=="svelte-1u5shki"&&(Ze.textContent=Rt),ht=r(X),je=c(X,"P",{"data-svelte-h":!0}),f(je)!=="svelte-1ffteuy"&&(je.innerHTML=Yt),gt=r(X),V=c(X,"DIV",{class:!0});var Ge=I(V);h(oe.$$.fragment,Ge),bt=r(Ge),We=c(Ge,"P",{"data-svelte-h":!0}),f(We)!=="svelte-50j04k"&&(We.textContent=Nt),Jt=r(Ge),h(H.$$.fragment,Ge),Ge.forEach(s),_t=r(X),E=c(X,"DIV",{class:!0});var Oe=I(E);h(le.$$.fragment,Oe),wt=r(Oe),ve=c(Oe,"P",{"data-svelte-h":!0}),f(ve)!=="svelte-16q0ax1"&&(ve.textContent=Ft),Oe.forEach(s),X.forEach(s),Se=r(e),h(ie.$$.fragment,e),Pe=r(e),x=c(e,"DIV",{class:!0});var et=I(x);h(re.$$.fragment,et),Tt=r(et),Ie=c(et,"P",{"data-svelte-h":!0}),f(Ie)!=="svelte-fkhkmf"&&(Ie.textContent=Ht),et.forEach(s),Ae=r(e),h(de.$$.fragment,e),Le=r(e),Ve=c(e,"P",{}),I(Ve).forEach(s),this.h()},h(){Q(n,"name","hf:doc:metadata"),Q(n,"content",rn),qt(l,"float","right"),Q(k,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Q(F,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Q(Z,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Q(V,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Q(E,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Q(j,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Q(x,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8")},m(e,a){o(document.head,n),u(e,w,a),u(e,m,a),u(e,d,a),u(e,l,a),u(e,M,a),g(U,e,a),u(e,T,a),u(e,B,a),u(e,xe,a),u(e,$,a),u(e,Qe,a),g(R,e,a),u(e,Re,a),u(e,z,a),u(e,Ye,a),g(Y,e,a),u(e,Ne,a),g(S,e,a),u(e,Fe,a),u(e,v,a),o(v,P),o(P,Me),o(P,tt),o(P,A),o(A,fe),o(A,nt),g(L,A,null),o(v,st),o(v,q),o(q,ye),o(q,at),o(q,D),o(D,he),o(D,ot),g(K,D,null),o(v,lt),o(v,ge),o(v,it),o(v,be),o(v,rt),o(v,Je),u(e,He,a),g(O,e,a),u(e,Ee,a),u(e,Z,a),g(ee,Z,null),o(Z,dt),o(Z,_e),o(Z,pt),o(Z,we),o(Z,ct),o(Z,k),g(te,k,null),o(k,mt),o(k,Te),o(k,ut),g(N,k,null),o(Z,Mt),o(Z,F),g(ne,F,null),o(F,ft),o(F,Ue),u(e,$e,a),g(se,e,a),u(e,ze,a),u(e,j,a),g(ae,j,null),o(j,yt),o(j,Ze),o(j,ht),o(j,je),o(j,gt),o(j,V),g(oe,V,null),o(V,bt),o(V,We),o(V,Jt),g(H,V,null),o(j,_t),o(j,E),g(le,E,null),o(E,wt),o(E,ve),u(e,Se,a),g(ie,e,a),u(e,Pe,a),u(e,x,a),g(re,x,null),o(x,Tt),o(x,Ie),u(e,Ae,a),g(de,e,a),u(e,Le,a),u(e,Ve,a),qe=!0},p(e,[a]){const G={};a&2&&(G.$$scope={dirty:a,ctx:e}),R.$set(G);const pe={};a&2&&(pe.$$scope={dirty:a,ctx:e}),Y.$set(pe);const ce={};a&2&&(ce.$$scope={dirty:a,ctx:e}),N.$set(ce);const me={};a&2&&(me.$$scope={dirty:a,ctx:e}),H.$set(me)},i(e){qe||(b(U.$$.fragment,e),b(R.$$.fragment,e),b(Y.$$.fragment,e),b(S.$$.fragment,e),b(L.$$.fragment,e),b(K.$$.fragment,e),b(O.$$.fragment,e),b(ee.$$.fragment,e),b(te.$$.fragment,e),b(N.$$.fragment,e),b(ne.$$.fragment,e),b(se.$$.fragment,e),b(ae.$$.fragment,e),b(oe.$$.fragment,e),b(H.$$.fragment,e),b(le.$$.fragment,e),b(ie.$$.fragment,e),b(re.$$.fragment,e),b(de.$$.fragment,e),qe=!0)},o(e){J(U.$$.fragment,e),J(R.$$.fragment,e),J(Y.$$.fragment,e),J(S.$$.fragment,e),J(L.$$.fragment,e),J(K.$$.fragment,e),J(O.$$.fragment,e),J(ee.$$.fragment,e),J(te.$$.fragment,e),J(N.$$.fragment,e),J(ne.$$.fragment,e),J(se.$$.fragment,e),J(ae.$$.fragment,e),J(oe.$$.fragment,e),J(H.$$.fragment,e),J(le.$$.fragment,e),J(ie.$$.fragment,e),J(re.$$.fragment,e),J(de.$$.fragment,e),qe=!1},d(e){e&&(s(w),s(m),s(d),s(l),s(M),s(T),s(B),s(xe),s($),s(Qe),s(Re),s(z),s(Ye),s(Ne),s(Fe),s(v),s(He),s(Ee),s(Z),s($e),s(ze),s(j),s(Se),s(Pe),s(x),s(Ae),s(Le),s(Ve)),s(n),_(U,e),_(R,e),_(Y,e),_(S,e),_(L),_(K),_(O,e),_(ee),_(te),_(N),_(ne),_(se,e),_(ae),_(oe),_(H),_(le),_(ie,e),_(re),_(de,e)}}}const rn='{"title":"Wan2.1","local":"wan21","sections":[{"title":"Notes","local":"notes","sections":[],"depth":2},{"title":"WanPipeline","local":"diffusers.WanPipeline","sections":[],"depth":2},{"title":"WanImageToVideoPipeline","local":"diffusers.WanImageToVideoPipeline","sections":[],"depth":2},{"title":"WanPipelineOutput","local":"diffusers.pipelines.wan.pipeline_output.WanPipelineOutput","sections":[],"depth":2}],"depth":1}';function dn(W){return St(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class gn extends Pt{constructor(n){super(),At(this,n,dn,ln,zt,{})}}export{gn as component};