Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,145 +1,145 @@
|
|
| 1 |
-
import gradio as gr
|
| 2 |
-
import os
|
| 3 |
-
import requests
|
| 4 |
-
import re
|
| 5 |
-
|
| 6 |
-
# Load GROQ API key from environment
|
| 7 |
-
GROQ_API_KEY = os.environ.get("GROQ_API_KEY")
|
| 8 |
-
GROQ_API_URL = "https://api.groq.com/openai/v1/chat/completions"
|
| 9 |
-
MODEL_NAME = "llama3-8b-8192"
|
| 10 |
-
|
| 11 |
-
SYSTEM_PROMPT = """You are 'ScriptForge AI', a professional YouTube Script Writer.
|
| 12 |
-
Your goal is to write highly engaging, audience-first scripts in the 2nd person (using 'You', 'Your').
|
| 13 |
-
Talk directly to the viewer as if you are the creator speaking to the camera.
|
| 14 |
-
|
| 15 |
-
FORMATTING RULES:
|
| 16 |
-
1. Every script must alternate between [SCENE DESCRIPTION] and [SCRIPT].
|
| 17 |
-
2. [SCENE DESCRIPTION] should describe the visuals, camera angles, or B-roll.
|
| 18 |
-
3. [SCRIPT] should be the actual spoken words.
|
| 19 |
-
4. Structure the script with a Hook, Intro, Main Points, and a Call to Action (CTA).
|
| 20 |
-
5. Maintain the requested Tone and Duration.
|
| 21 |
-
|
| 22 |
-
Example:
|
| 23 |
-
[SCENE DESCRIPTION]: Close-up of the product.
|
| 24 |
-
[SCRIPT]: You need to see this to believe it.
|
| 25 |
-
"""
|
| 26 |
-
|
| 27 |
-
def parse_script(full_text):
|
| 28 |
-
# Extract [SCRIPT] parts for TTS
|
| 29 |
-
script_parts = re.findall(r'\[SCRIPT\]:(.*?)(?=\[SCENE DESCRIPTION\]|$)', full_text, re.DOTALL | re.IGNORECASE)
|
| 30 |
-
clean_script = "\n".join([p.strip() for p in script_parts])
|
| 31 |
-
|
| 32 |
-
# Extract [SCENE DESCRIPTION] parts for Visuals
|
| 33 |
-
scene_parts = re.findall(r'\[SCENE DESCRIPTION\]:(.*?)(?=\[SCRIPT\]|$)', full_text, re.DOTALL | re.IGNORECASE)
|
| 34 |
-
clean_scenes = "\n".join([p.strip() for p in scene_parts])
|
| 35 |
-
|
| 36 |
-
# Calculate stats
|
| 37 |
-
word_count = len(clean_script.split())
|
| 38 |
-
# Est duration: ~150 words per minute
|
| 39 |
-
duration_minutes = word_count / 150
|
| 40 |
-
minutes = int(duration_minutes)
|
| 41 |
-
seconds = int((duration_minutes - minutes) * 60)
|
| 42 |
-
duration_str = f"{minutes}m {seconds}s"
|
| 43 |
-
|
| 44 |
-
return clean_script, clean_scenes, word_count, duration_str
|
| 45 |
-
|
| 46 |
-
def query_groq(topic, tone, duration, hook_strength, chat_history):
|
| 47 |
-
if not GROQ_API_KEY:
|
| 48 |
-
return "Error: GROQ_API_KEY not found in environment secrets. Please add it in Settings > Secrets.", "", "", 0, "0m 0s"
|
| 49 |
-
|
| 50 |
-
headers = {
|
| 51 |
-
"Authorization": f"Bearer {GROQ_API_KEY}",
|
| 52 |
-
"Content-Type": "application/json"
|
| 53 |
-
}
|
| 54 |
-
|
| 55 |
-
user_input = f"Topic: {topic}\nTone: {tone}\nTarget Duration: {duration}\nAction: Write a full YouTube script."
|
| 56 |
-
|
| 57 |
-
messages = [{"role": "system", "content": SYSTEM_PROMPT}]
|
| 58 |
-
# Add history for context (optional but helpful)
|
| 59 |
-
for user, bot in chat_history[-3:]: # Keep last 3 exchanges to avoid context bloat
|
| 60 |
-
messages.append({"role": "user", "content": user})
|
| 61 |
-
messages.append({"role": "assistant", "content": bot})
|
| 62 |
-
|
| 63 |
-
messages.append({"role": "user", "content": user_input})
|
| 64 |
-
|
| 65 |
-
try:
|
| 66 |
-
response = requests.post(GROQ_API_URL, headers=headers, json={
|
| 67 |
-
"model": MODEL_NAME,
|
| 68 |
-
"messages": messages,
|
| 69 |
-
"temperature": hook_strength
|
| 70 |
-
}, timeout=30)
|
| 71 |
-
|
| 72 |
-
if response.status_code == 200:
|
| 73 |
-
full_reply = response.json()["choices"][0]["message"]["content"]
|
| 74 |
-
tts_script, scenes, wc, dur = parse_script(full_reply)
|
| 75 |
-
return full_reply, tts_script, scenes, wc, dur
|
| 76 |
-
else:
|
| 77 |
-
return f"Error {response.status_code}: {response.text}", "", "", 0, "0m 0s"
|
| 78 |
-
except Exception as e:
|
| 79 |
-
return f"Request failed: {str(e)}", "", "", 0, "0m 0s"
|
| 80 |
-
|
| 81 |
-
def respond(topic, tone, duration, hook_strength, chat_history):
|
| 82 |
-
full_reply, tts_script, scenes, wc, dur = query_groq(topic, tone, duration, hook_strength, chat_history)
|
| 83 |
-
chat_history.append((topic, full_reply))
|
| 84 |
-
return chat_history, tts_script, scenes, f"{wc} words", dur
|
| 85 |
-
|
| 86 |
-
css = """
|
| 87 |
-
footer {visibility: hidden}
|
| 88 |
-
.stat-box {
|
| 89 |
-
background-color: #f0f2f6;
|
| 90 |
-
padding: 10px;
|
| 91 |
-
border-radius: 10px;
|
| 92 |
-
text-align: center;
|
| 93 |
-
border: 1px solid #ddd;
|
| 94 |
-
}
|
| 95 |
-
"""
|
| 96 |
-
|
| 97 |
-
with gr.Blocks(
|
| 98 |
-
gr.Markdown("# 🎬 ScriptForge AI: YouTube Script Master")
|
| 99 |
-
gr.Markdown("Transform your video ideas into high-retention, audience-first scripts. *Powered by GROQ*")
|
| 100 |
-
|
| 101 |
-
with gr.Row():
|
| 102 |
-
with gr.Column(scale=1):
|
| 103 |
-
topic = gr.Textbox(label="Video Topic/Description", placeholder="E.g., How to build a PC in 2025", lines=3)
|
| 104 |
-
tone = gr.Dropdown(
|
| 105 |
-
choices=["High Energy", "Storytelling", "Educational", "Minimalist", "Aggressive/Hype"],
|
| 106 |
-
label="Video Tone",
|
| 107 |
-
value="High Energy"
|
| 108 |
-
)
|
| 109 |
-
duration = gr.Dropdown(
|
| 110 |
-
choices=["Shorts (<60s)", "Standard (5-10 mins)", "Deep Dive (15+ mins)"],
|
| 111 |
-
label="Target Duration",
|
| 112 |
-
value="Standard (5-10 mins)"
|
| 113 |
-
)
|
| 114 |
-
hook_strength = gr.Slider(minimum=0.1, maximum=1.5, value=0.7, step=0.1, label="Hook Strength (Creativity)")
|
| 115 |
-
generate_btn = gr.Button("🚀 Generate Script", variant="primary")
|
| 116 |
-
clear = gr.Button("Clear")
|
| 117 |
-
|
| 118 |
-
with gr.Column(scale=2):
|
| 119 |
-
with gr.Row():
|
| 120 |
-
word_count_display = gr.Textbox(label="Word Count", interactive=False, elem_classes="stat-box")
|
| 121 |
-
duration_display = gr.Textbox(label="Est. Speaking Time", interactive=False, elem_classes="stat-box")
|
| 122 |
-
|
| 123 |
-
with gr.Tabs():
|
| 124 |
-
with gr.TabItem("Combined View"):
|
| 125 |
-
chatbot = gr.Chatbot(height=500)
|
| 126 |
-
|
| 127 |
-
with gr.TabItem("TTS Only (Dialogue)"):
|
| 128 |
-
tts_output = gr.Textbox(label="Copy this for Text-to-Speech",
|
| 129 |
-
|
| 130 |
-
with gr.TabItem("Visuals Only (Shot List)"):
|
| 131 |
-
scenes_output = gr.Textbox(label="Video Scene Descriptions", lines=20)
|
| 132 |
-
|
| 133 |
-
# State for history
|
| 134 |
-
state = gr.State([])
|
| 135 |
-
|
| 136 |
-
generate_btn.click(
|
| 137 |
-
respond,
|
| 138 |
-
[topic, tone, duration, hook_strength, state],
|
| 139 |
-
[chatbot, tts_output, scenes_output, word_count_display, duration_display]
|
| 140 |
-
)
|
| 141 |
-
|
| 142 |
-
clear.click(lambda: (None, "", "", "", "", []), None, [topic, chatbot, tts_output, scenes_output, word_count_display, duration_display])
|
| 143 |
-
|
| 144 |
-
if __name__ == "__main__":
|
| 145 |
-
demo.launch()
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
import os
|
| 3 |
+
import requests
|
| 4 |
+
import re
|
| 5 |
+
|
| 6 |
+
# Load GROQ API key from environment
|
| 7 |
+
GROQ_API_KEY = os.environ.get("GROQ_API_KEY")
|
| 8 |
+
GROQ_API_URL = "https://api.groq.com/openai/v1/chat/completions"
|
| 9 |
+
MODEL_NAME = "llama3-8b-8192"
|
| 10 |
+
|
| 11 |
+
SYSTEM_PROMPT = """You are 'ScriptForge AI', a professional YouTube Script Writer.
|
| 12 |
+
Your goal is to write highly engaging, audience-first scripts in the 2nd person (using 'You', 'Your').
|
| 13 |
+
Talk directly to the viewer as if you are the creator speaking to the camera.
|
| 14 |
+
|
| 15 |
+
FORMATTING RULES:
|
| 16 |
+
1. Every script must alternate between [SCENE DESCRIPTION] and [SCRIPT].
|
| 17 |
+
2. [SCENE DESCRIPTION] should describe the visuals, camera angles, or B-roll.
|
| 18 |
+
3. [SCRIPT] should be the actual spoken words.
|
| 19 |
+
4. Structure the script with a Hook, Intro, Main Points, and a Call to Action (CTA).
|
| 20 |
+
5. Maintain the requested Tone and Duration.
|
| 21 |
+
|
| 22 |
+
Example:
|
| 23 |
+
[SCENE DESCRIPTION]: Close-up of the product.
|
| 24 |
+
[SCRIPT]: You need to see this to believe it.
|
| 25 |
+
"""
|
| 26 |
+
|
| 27 |
+
def parse_script(full_text):
|
| 28 |
+
# Extract [SCRIPT] parts for TTS
|
| 29 |
+
script_parts = re.findall(r'\[SCRIPT\]:(.*?)(?=\[SCENE DESCRIPTION\]|$)', full_text, re.DOTALL | re.IGNORECASE)
|
| 30 |
+
clean_script = "\n".join([p.strip() for p in script_parts])
|
| 31 |
+
|
| 32 |
+
# Extract [SCENE DESCRIPTION] parts for Visuals
|
| 33 |
+
scene_parts = re.findall(r'\[SCENE DESCRIPTION\]:(.*?)(?=\[SCRIPT\]|$)', full_text, re.DOTALL | re.IGNORECASE)
|
| 34 |
+
clean_scenes = "\n".join([p.strip() for p in scene_parts])
|
| 35 |
+
|
| 36 |
+
# Calculate stats
|
| 37 |
+
word_count = len(clean_script.split())
|
| 38 |
+
# Est duration: ~150 words per minute
|
| 39 |
+
duration_minutes = word_count / 150
|
| 40 |
+
minutes = int(duration_minutes)
|
| 41 |
+
seconds = int((duration_minutes - minutes) * 60)
|
| 42 |
+
duration_str = f"{minutes}m {seconds}s"
|
| 43 |
+
|
| 44 |
+
return clean_script, clean_scenes, word_count, duration_str
|
| 45 |
+
|
| 46 |
+
def query_groq(topic, tone, duration, hook_strength, chat_history):
|
| 47 |
+
if not GROQ_API_KEY:
|
| 48 |
+
return "Error: GROQ_API_KEY not found in environment secrets. Please add it in Settings > Secrets.", "", "", 0, "0m 0s"
|
| 49 |
+
|
| 50 |
+
headers = {
|
| 51 |
+
"Authorization": f"Bearer {GROQ_API_KEY}",
|
| 52 |
+
"Content-Type": "application/json"
|
| 53 |
+
}
|
| 54 |
+
|
| 55 |
+
user_input = f"Topic: {topic}\nTone: {tone}\nTarget Duration: {duration}\nAction: Write a full YouTube script."
|
| 56 |
+
|
| 57 |
+
messages = [{"role": "system", "content": SYSTEM_PROMPT}]
|
| 58 |
+
# Add history for context (optional but helpful)
|
| 59 |
+
for user, bot in chat_history[-3:]: # Keep last 3 exchanges to avoid context bloat
|
| 60 |
+
messages.append({"role": "user", "content": user})
|
| 61 |
+
messages.append({"role": "assistant", "content": bot})
|
| 62 |
+
|
| 63 |
+
messages.append({"role": "user", "content": user_input})
|
| 64 |
+
|
| 65 |
+
try:
|
| 66 |
+
response = requests.post(GROQ_API_URL, headers=headers, json={
|
| 67 |
+
"model": MODEL_NAME,
|
| 68 |
+
"messages": messages,
|
| 69 |
+
"temperature": hook_strength
|
| 70 |
+
}, timeout=30)
|
| 71 |
+
|
| 72 |
+
if response.status_code == 200:
|
| 73 |
+
full_reply = response.json()["choices"][0]["message"]["content"]
|
| 74 |
+
tts_script, scenes, wc, dur = parse_script(full_reply)
|
| 75 |
+
return full_reply, tts_script, scenes, wc, dur
|
| 76 |
+
else:
|
| 77 |
+
return f"Error {response.status_code}: {response.text}", "", "", 0, "0m 0s"
|
| 78 |
+
except Exception as e:
|
| 79 |
+
return f"Request failed: {str(e)}", "", "", 0, "0m 0s"
|
| 80 |
+
|
| 81 |
+
def respond(topic, tone, duration, hook_strength, chat_history):
|
| 82 |
+
full_reply, tts_script, scenes, wc, dur = query_groq(topic, tone, duration, hook_strength, chat_history)
|
| 83 |
+
chat_history.append((topic, full_reply))
|
| 84 |
+
return chat_history, tts_script, scenes, f"{wc} words", dur
|
| 85 |
+
|
| 86 |
+
css = """
|
| 87 |
+
footer {visibility: hidden}
|
| 88 |
+
.stat-box {
|
| 89 |
+
background-color: #f0f2f6;
|
| 90 |
+
padding: 10px;
|
| 91 |
+
border-radius: 10px;
|
| 92 |
+
text-align: center;
|
| 93 |
+
border: 1px solid #ddd;
|
| 94 |
+
}
|
| 95 |
+
"""
|
| 96 |
+
|
| 97 |
+
with gr.Blocks() as demo:
|
| 98 |
+
gr.Markdown("# 🎬 ScriptForge AI: YouTube Script Master")
|
| 99 |
+
gr.Markdown("Transform your video ideas into high-retention, audience-first scripts. *Powered by GROQ*")
|
| 100 |
+
|
| 101 |
+
with gr.Row():
|
| 102 |
+
with gr.Column(scale=1):
|
| 103 |
+
topic = gr.Textbox(label="Video Topic/Description", placeholder="E.g., How to build a PC in 2025", lines=3)
|
| 104 |
+
tone = gr.Dropdown(
|
| 105 |
+
choices=["High Energy", "Storytelling", "Educational", "Minimalist", "Aggressive/Hype"],
|
| 106 |
+
label="Video Tone",
|
| 107 |
+
value="High Energy"
|
| 108 |
+
)
|
| 109 |
+
duration = gr.Dropdown(
|
| 110 |
+
choices=["Shorts (<60s)", "Standard (5-10 mins)", "Deep Dive (15+ mins)"],
|
| 111 |
+
label="Target Duration",
|
| 112 |
+
value="Standard (5-10 mins)"
|
| 113 |
+
)
|
| 114 |
+
hook_strength = gr.Slider(minimum=0.1, maximum=1.5, value=0.7, step=0.1, label="Hook Strength (Creativity)")
|
| 115 |
+
generate_btn = gr.Button("🚀 Generate Script", variant="primary")
|
| 116 |
+
clear = gr.Button("Clear")
|
| 117 |
+
|
| 118 |
+
with gr.Column(scale=2):
|
| 119 |
+
with gr.Row():
|
| 120 |
+
word_count_display = gr.Textbox(label="Word Count", interactive=False, elem_classes="stat-box")
|
| 121 |
+
duration_display = gr.Textbox(label="Est. Speaking Time", interactive=False, elem_classes="stat-box")
|
| 122 |
+
|
| 123 |
+
with gr.Tabs():
|
| 124 |
+
with gr.TabItem("Combined View"):
|
| 125 |
+
chatbot = gr.Chatbot(height=500)
|
| 126 |
+
|
| 127 |
+
with gr.TabItem("TTS Only (Dialogue)"):
|
| 128 |
+
tts_output = gr.Textbox(label="Copy this for Text-to-Speech", lines=20)
|
| 129 |
+
|
| 130 |
+
with gr.TabItem("Visuals Only (Shot List)"):
|
| 131 |
+
scenes_output = gr.Textbox(label="Video Scene Descriptions", lines=20)
|
| 132 |
+
|
| 133 |
+
# State for history
|
| 134 |
+
state = gr.State([])
|
| 135 |
+
|
| 136 |
+
generate_btn.click(
|
| 137 |
+
respond,
|
| 138 |
+
[topic, tone, duration, hook_strength, state],
|
| 139 |
+
[chatbot, tts_output, scenes_output, word_count_display, duration_display]
|
| 140 |
+
)
|
| 141 |
+
|
| 142 |
+
clear.click(lambda: (None, "", "", "", "", []), None, [topic, chatbot, tts_output, scenes_output, word_count_display, duration_display])
|
| 143 |
+
|
| 144 |
+
if __name__ == "__main__":
|
| 145 |
+
demo.launch(theme=gr.themes.Soft(), css=css)
|