kadirnar commited on
Commit
d8f08b3
1 Parent(s): 29635c0

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +13 -3
app.py CHANGED
@@ -6,6 +6,7 @@ import gradio as gr
6
 
7
 
8
  def tune_video_predict(
 
9
  prompt: str,
10
  video_length: int,
11
  height: int,
@@ -13,7 +14,7 @@ def tune_video_predict(
13
  num_inference_steps: int,
14
  guidance_scale: float,
15
  ):
16
- unet = UNet3DConditionModel.from_pretrained('Tune-A-Video-library/a-man-is-surfing', subfolder='unet', torch_dtype=torch.float16).to('cuda')
17
  pipe = TuneAVideoPipeline.from_pretrained('CompVis/stable-diffusion-v1-4', unet=unet, torch_dtype=torch.float16).to("cuda")
18
  video = pipe(prompt, video_length=video_length, height=height, width=width, num_inference_steps=num_inference_steps, guidance_scale=guidance_scale).videos
19
  output_path = save_videos_grid(video, save_path='output', path=f"{prompt}.gif")
@@ -21,6 +22,13 @@ def tune_video_predict(
21
 
22
 
23
  demo_inputs = [
 
 
 
 
 
 
 
24
  gr.inputs.Textbox(
25
  label="Prompt",
26
  default='a flower blooming'
@@ -67,8 +75,10 @@ demo_inputs = [
67
  demo_outputs = gr.outputs.Video(type="gif", label="Output")
68
 
69
  examples = [
70
- ["a panda is surfing", 8, 416, 416, 50, 7.5],
71
- ["a flower blooming", 5, 416, 416, 50, 7.5],
 
 
72
  ]
73
 
74
  description = "This generates video from an input text, using [one-shot tuning of diffusion models](https://arxiv.org/abs/2212.11565). To use it, simply input a text."
 
6
 
7
 
8
  def tune_video_predict(
9
+ pipe_id: str,
10
  prompt: str,
11
  video_length: int,
12
  height: int,
 
14
  num_inference_steps: int,
15
  guidance_scale: float,
16
  ):
17
+ unet = UNet3DConditionModel.from_pretrained(pipe_id, subfolder='unet', torch_dtype=torch.float16).to('cuda')
18
  pipe = TuneAVideoPipeline.from_pretrained('CompVis/stable-diffusion-v1-4', unet=unet, torch_dtype=torch.float16).to("cuda")
19
  video = pipe(prompt, video_length=video_length, height=height, width=width, num_inference_steps=num_inference_steps, guidance_scale=guidance_scale).videos
20
  output_path = save_videos_grid(video, save_path='output', path=f"{prompt}.gif")
 
22
 
23
 
24
  demo_inputs = [
25
+ gr.inputs.Dropdown(
26
+ label="Model",
27
+ choices=[
28
+ "Tune-A-Video-library/a-man-is-surfing",
29
+ "sd-dreambooth-library/mr-potato-head",
30
+ ]
31
+ ),
32
  gr.inputs.Textbox(
33
  label="Prompt",
34
  default='a flower blooming'
 
75
  demo_outputs = gr.outputs.Video(type="gif", label="Output")
76
 
77
  examples = [
78
+ ["Tune-A-Video-library/a-man-is-surfing", "a panda is surfing", 5, 416, 416, 50, 7.5],
79
+ ["Tune-A-Video-library/a-man-is-surfing", "a flower blooming", 5, 416, 416, 50, 7.5],
80
+ ["sd-dreambooth-library/mr-potato-head", "sks mr potato head, wearing a pink hat, is surfing.", 5, 416, 416, 50, 7.5],
81
+ ["sd-dreambooth-library/mr-potato-head", "sks mr potato head is surfing in the forest.", 5, 416, 416, 50, 7.5],
82
  ]
83
 
84
  description = "This generates video from an input text, using [one-shot tuning of diffusion models](https://arxiv.org/abs/2212.11565). To use it, simply input a text."