Update app.py

app.py

@@ -1,25 +1,32 @@
 import gradio as gr
-from pydub import AudioSegment, silence
+import requests
+import random
+import urllib.parse
+import tempfile
+import os
 import nltk
 import srt
-import io
+from pydub import AudioSegment, silence
 import datetime
 
-nltk.download('punkt')
+nltk.download("punkt")
 
-def process_audio_and_script(audio_file, script_text):
-    # Load audio
-    audio = AudioSegment.from_file(audio_file)
-    silence_thresh = audio.dBFS - 16
-    silences = silence.detect_silence(audio, min_silence_len=400, silence_thresh=silence_thresh)
+NSFW_URL_TEMPLATE = os.getenv("NSFW_API_URL_TEMPLATE")
+TTS_URL_TEMPLATE = os.getenv("TTS_API_URL_TEMPLATE")
 
-    # Convert silence list to start-end in seconds
-    silences = [(start / 1000, stop / 1000) for start, stop in silences]
+if not NSFW_URL_TEMPLATE:
+    raise ValueError("Missing Secret: NSFW_API_URL_TEMPLATE is not set.")
+if not TTS_URL_TEMPLATE:
+    raise ValueError("Missing Secret: TTS_API_URL_TEMPLATE is not set.")
 
-    # Segment script based on punctuation
+VOICES = ["alloy", "echo", "fable", "onyx", "nova", "shimmer", "coral", "verse", "ballad", "ash", "sage", "amuch", "dan"]
+
+def generate_srt(audio_path, script_text):
+    audio = AudioSegment.from_file(audio_path)
+    silences = silence.detect_silence(audio, min_silence_len=300, silence_thresh=audio.dBFS - 16)
+    silences = [(start / 1000.0, end / 1000.0) for start, end in silences]
     sentences = nltk.tokenize.sent_tokenize(script_text)
 
-    # Distribute timing across sentences based on silence gaps
     subtitles = []
     last_time = 0.0
     for i, sentence in enumerate(sentences):
@@ -29,55 +36,98 @@ def process_audio_and_script(audio_file, script_text):
             last_time = silences[i][1]
         else:
             start = last_time
-            end = start + 2.5  # default length if not enough silences
-
-        subtitle = srt.Subtitle(index=i + 1,
-                                start=datetime.timedelta(seconds=start),
-                                end=datetime.timedelta(seconds=end),
-                                content=sentence)
+            end = start + 2.5  # default
+        subtitle = srt.Subtitle(
+            index=i + 1,
+            start=datetime.timedelta(seconds=start),
+            end=datetime.timedelta(seconds=end),
+            content=sentence
+        )
         subtitles.append(subtitle)
 
-    srt_file = srt.compose(subtitles)
-    return srt_file
+    srt_data = srt.compose(subtitles)
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".srt", mode='w') as srt_file:
+        srt_file.write(srt_data)
+        return srt_file.name
+
+def check_nsfw(prompt: str) -> bool:
+    encoded_prompt = urllib.parse.quote(prompt)
+    url = NSFW_URL_TEMPLATE.format(prompt=encoded_prompt)
+    try:
+        response = requests.get(url, timeout=20)
+        result = response.text.strip().upper()
+        return result == "YES"
+    except:
+        raise gr.Error("Failed to check prompt safety.")
+
+def generate_audio(prompt: str, voice: str, emotion: str, seed: int) -> bytes:
+    encoded_prompt = urllib.parse.quote(prompt)
+    encoded_emotion = urllib.parse.quote(emotion)
+    url = TTS_URL_TEMPLATE.format(prompt=encoded_prompt, emotion=encoded_emotion, voice=voice, seed=seed)
+    response = requests.get(url, timeout=60)
+    if 'audio' not in response.headers.get('content-type', ''):
+        raise gr.Error("Invalid audio response.")
+    return response.content
+
+def text_to_speech_app(prompt, voice, emotion, use_random_seed, specific_seed, subtitle_script):
+    if not prompt or not voice:
+        raise gr.Error("Prompt and Voice are required.")
+    seed = random.randint(0, 2**32 - 1) if use_random_seed else int(specific_seed)
+    is_nsfw = False  # You can enable this by: is_nsfw = check_nsfw(prompt)
 
-def download_srt(audio_file, script_text):
-    srt_data = process_audio_and_script(audio_file, script_text)
-    return ("subtitles.srt", srt_data)
+    if is_nsfw:
+        return None, None, "Prompt is flagged NSFW"
+
+    try:
+        audio_bytes = generate_audio(prompt, voice, emotion, seed)
+        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as temp_audio:
+            temp_audio.write(audio_bytes)
+            audio_path = temp_audio.name
+
+        srt_path = None
+        if subtitle_script.strip():
+            srt_path = generate_srt(audio_path, subtitle_script)
+
+        return audio_path, srt_path, f"Audio and SRT generated with seed {seed}"
+    except Exception as e:
+        return None, None, f"Error: {str(e)}"
+
+def toggle_seed_input(use_random_seed):
+    return gr.update(visible=not use_random_seed, value=12345)
+
+with gr.Blocks() as app:
+    gr.Markdown("## 🎙️ Advanced OpenAI TTS + Subtitle Generator")
 
-# Interface
-with gr.Blocks() as demo:
-    gr.Markdown("### 🎙️ Audio to Timed Subtitle (SRT) Generator with Waveform")
     with gr.Row():
-        audio_input = gr.Audio(type="file", label="Upload Audio")
-        script_input = gr.Textbox(lines=10, label="Paste Script/Text with Punctuation")
-    
-    srt_output = gr.File(label="Download SRT")
-
-    def waveform_html(audio_file):
-        return f"""
-        <div id="waveform"></div>
-        <script src="https://unpkg.com/wavesurfer.js"></script>
-        <script>
-        var wavesurfer = WaveSurfer.create({
-            container: '#waveform',
-            waveColor: '#999',
-            progressColor: '#333',
-            height: 100
-        });
-        wavesurfer.load("{audio_file}");
-        </script>
-        """
-    
-    waveform = gr.HTML()
+        with gr.Column(scale=2):
+            prompt_input = gr.Textbox(label="Prompt", placeholder="Enter your text...")
+            emotion_input = gr.Textbox(label="Emotion Style", placeholder="happy, sad, excited, calm...")
+            voice_dropdown = gr.Dropdown(label="Voice", choices=VOICES, value="alloy")
+            subtitle_script = gr.Textbox(label="Subtitle Script", lines=6, placeholder="Paste script here for SRT generation")
+        with gr.Column(scale=1):
+            random_seed_checkbox = gr.Checkbox(label="Use Random Seed", value=True)
+            seed_input = gr.Number(label="Specific Seed", value=12345, visible=False, precision=0)
+
+    submit_button = gr.Button("🎧 Generate Audio + Subtitles", variant="primary")
 
     with gr.Row():
-        gen_btn = gr.Button("Generate SRT")
-        gen_btn.click(fn=download_srt,
-                      inputs=[audio_input, script_input],
-                      outputs=srt_output)
+        audio_output = gr.Audio(label="Generated Audio", type="filepath")
+        srt_output = gr.File(label="Download SRT File")
+        status_output = gr.Textbox(label="Status")
+
+    random_seed_checkbox.change(
+        fn=toggle_seed_input,
+        inputs=[random_seed_checkbox],
+        outputs=[seed_input]
+    )
 
-        audio_input.change(fn=lambda audio: waveform_html(audio["name"]),
-                           inputs=audio_input,
-                           outputs=waveform)
+    submit_button.click(
+        fn=text_to_speech_app,
+        inputs=[prompt_input, voice_dropdown, emotion_input, random_seed_checkbox, seed_input, subtitle_script],
+        outputs=[audio_output, srt_output, status_output],
+        concurrency_limit=30
+    )
 
-demo.launch()
+if __name__ == "__main__":
+    if NSFW_URL_TEMPLATE and TTS_URL_TEMPLATE:
+        app.launch()