Upload 3 files
Browse files- app.py +8 -7
- img_gen.py +69 -0
- prompt_generation.py +1 -0
app.py
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
import streamlit as st
|
2 |
from gtts import gTTS
|
3 |
|
4 |
-
from
|
5 |
from prompt_generation import pipeline
|
6 |
|
7 |
|
@@ -15,7 +15,7 @@ def page_navigation(current_page):
|
|
15 |
current_page -= 1
|
16 |
|
17 |
with col2:
|
18 |
-
|
19 |
|
20 |
if current_page < 10:
|
21 |
with col3:
|
@@ -23,10 +23,11 @@ def page_navigation(current_page):
|
|
23 |
if current_page == 0:
|
24 |
user_input = st.session_state.user_input
|
25 |
prompt_response = pipeline(user_input, 10)
|
26 |
-
|
27 |
init_prompt = prompt_response.get("story")
|
28 |
|
29 |
-
init_img, img_dict = generate_story(init_prompt,
|
|
|
30 |
|
31 |
st.session_state.pipeline_response = prompt_response
|
32 |
st.session_state.init_img = init_img
|
@@ -42,7 +43,7 @@ def get_pipeline_data(page_number):
|
|
42 |
pipeline_response = st.session_state.pipeline_response
|
43 |
text_output = pipeline_response.get("steps")[page_number - 1]
|
44 |
img_dict = st.session_state.img_dict
|
45 |
-
img = img_dict[page_number-1].get("image")
|
46 |
|
47 |
return {"text_output": text_output, "image_obj": img}
|
48 |
|
@@ -56,7 +57,7 @@ def main():
|
|
56 |
|
57 |
# Display content for each page
|
58 |
if current_page == 0:
|
59 |
-
st.write("
|
60 |
user_input = st.text_area("")
|
61 |
st.session_state.user_input = user_input
|
62 |
|
@@ -69,7 +70,7 @@ def main():
|
|
69 |
# Display text output
|
70 |
st.write(text_output)
|
71 |
|
72 |
-
tts = gTTS(text_output)
|
73 |
tts.save('audio.mp3')
|
74 |
st.audio('audio.mp3')
|
75 |
|
|
|
1 |
import streamlit as st
|
2 |
from gtts import gTTS
|
3 |
|
4 |
+
from img_gen import generate_story
|
5 |
from prompt_generation import pipeline
|
6 |
|
7 |
|
|
|
15 |
current_page -= 1
|
16 |
|
17 |
with col2:
|
18 |
+
print(f'Step {current_page} of 10')
|
19 |
|
20 |
if current_page < 10:
|
21 |
with col3:
|
|
|
23 |
if current_page == 0:
|
24 |
user_input = st.session_state.user_input
|
25 |
prompt_response = pipeline(user_input, 10)
|
26 |
+
image_prompts_steps = prompt_response.get("image_prompts_steps")
|
27 |
init_prompt = prompt_response.get("story")
|
28 |
|
29 |
+
init_img, img_dict = generate_story(init_prompt,
|
30 |
+
image_prompts_steps)
|
31 |
|
32 |
st.session_state.pipeline_response = prompt_response
|
33 |
st.session_state.init_img = init_img
|
|
|
43 |
pipeline_response = st.session_state.pipeline_response
|
44 |
text_output = pipeline_response.get("steps")[page_number - 1]
|
45 |
img_dict = st.session_state.img_dict
|
46 |
+
img = img_dict[page_number - 1].get("image")
|
47 |
|
48 |
return {"text_output": text_output, "image_obj": img}
|
49 |
|
|
|
57 |
|
58 |
# Display content for each page
|
59 |
if current_page == 0:
|
60 |
+
st.write("Describe a story you would like me to tell:")
|
61 |
user_input = st.text_area("")
|
62 |
st.session_state.user_input = user_input
|
63 |
|
|
|
70 |
# Display text output
|
71 |
st.write(text_output)
|
72 |
|
73 |
+
tts = gTTS(text_output.split(".", 1)[1])
|
74 |
tts.save('audio.mp3')
|
75 |
st.audio('audio.mp3')
|
76 |
|
img_gen.py
ADDED
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
|
3 |
+
from diffusers import StableDiffusionImg2ImgPipeline, \
|
4 |
+
StableDiffusionPipeline
|
5 |
+
|
6 |
+
|
7 |
+
def check_cuda_device():
|
8 |
+
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
9 |
+
return device
|
10 |
+
|
11 |
+
|
12 |
+
def get_the_model(device=None):
|
13 |
+
model_id = "stabilityai/stable-diffusion-2"
|
14 |
+
pipe = StableDiffusionPipeline.from_pretrained(model_id,
|
15 |
+
torch_dtype=torch.float16)
|
16 |
+
if device:
|
17 |
+
pipe.to(device)
|
18 |
+
else:
|
19 |
+
device = check_cuda_device()
|
20 |
+
pipe.to(device)
|
21 |
+
|
22 |
+
return pipe
|
23 |
+
|
24 |
+
|
25 |
+
def get_image_to_image_model(path=None, device=None):
|
26 |
+
model_id = "stabilityai/stable-diffusion-2"
|
27 |
+
if path:
|
28 |
+
pipe = StableDiffusionImg2ImgPipeline.from_pretrained(
|
29 |
+
path,
|
30 |
+
torch_dtype=torch.float16)
|
31 |
+
else:
|
32 |
+
pipe = StableDiffusionImg2ImgPipeline.from_pretrained(
|
33 |
+
model_id,
|
34 |
+
torch_dtype=torch.float16)
|
35 |
+
if device:
|
36 |
+
if device == "cuda" or device == "cpu":
|
37 |
+
pipe.to(device)
|
38 |
+
else:
|
39 |
+
device = check_cuda_device()
|
40 |
+
pipe.to(device)
|
41 |
+
|
42 |
+
return pipe
|
43 |
+
|
44 |
+
|
45 |
+
def gen_initial_img(int_prompt):
|
46 |
+
model = get_the_model(None)
|
47 |
+
image = model(int_prompt, num_inference_steps=100).images[0]
|
48 |
+
|
49 |
+
return image
|
50 |
+
|
51 |
+
|
52 |
+
def generate_story(int_prompt, steps, iterations=133):
|
53 |
+
image_dic = {}
|
54 |
+
init_img = gen_initial_img(int_prompt)
|
55 |
+
img2img_model = get_image_to_image_model()
|
56 |
+
img = init_img
|
57 |
+
|
58 |
+
for idx, step in enumerate(steps):
|
59 |
+
print(f"step: {idx}")
|
60 |
+
print(step)
|
61 |
+
image = img2img_model(prompt=step, image=img, strength=0.75, guidance_scale=7.5,
|
62 |
+
num_inference_steps=iterations).images[0]
|
63 |
+
image_dic[idx] = {
|
64 |
+
"image": image,
|
65 |
+
"prompt": step
|
66 |
+
}
|
67 |
+
img = image
|
68 |
+
|
69 |
+
return init_img, image_dic
|
prompt_generation.py
CHANGED
@@ -97,6 +97,7 @@ def pipeline(user_description: str, n_steps: int = 10) -> dict:
|
|
97 |
|
98 |
image_prompts = [fut.result() for fut in image_prompts_futures]
|
99 |
|
|
|
100 |
return {"story": story, "steps": steps, "image_prompts": image_prompts}
|
101 |
|
102 |
|
|
|
97 |
|
98 |
image_prompts = [fut.result() for fut in image_prompts_futures]
|
99 |
|
100 |
+
print(story)
|
101 |
return {"story": story, "steps": steps, "image_prompts": image_prompts}
|
102 |
|
103 |
|