File size: 18,525 Bytes
9db7bb3
2720196
043971a
 
 
 
 
2720196
 
043971a
2720196
2114e35
2720196
2114e35
cda1077
2720196
 
2114e35
eaeaf35
caca082
eaeaf35
2114e35
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
eaeaf35
 
 
2114e35
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2720196
 
eaeaf35
2114e35
cda1077
2114e35
 
 
 
7faf9f3
2114e35
 
 
 
2720196
2114e35
 
2720196
2114e35
2720196
2114e35
2720196
eaeaf35
4c6c365
 
2720196
2114e35
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2720196
2114e35
7888a85
2114e35
 
2720196
2114e35
 
2720196
2114e35
 
2720196
2114e35
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
eaeaf35
4c6c365
2114e35
4c6c365
 
2720196
2114e35
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4c6c365
 
2114e35
2720196
 
 
2114e35
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2720196
2114e35
 
 
 
 
 
4c6c365
f1ad00c
2114e35
4c6c365
2720196
 
2114e35
 
 
2720196
2114e35
 
 
 
 
 
 
 
 
2720196
2114e35
 
 
 
 
 
 
 
 
 
 
 
 
 
2720196
2114e35
 
 
 
4c6c365
2114e35
 
 
 
 
 
2720196
2114e35
 
 
 
 
 
 
 
 
 
 
 
 
 
2720196
2114e35
 
 
2720196
4c6c365
2720196
 
 
2114e35
 
 
 
 
 
 
 
 
2720196
 
2114e35
 
 
2720196
 
2114e35
 
 
4c6c365
2720196
2114e35
 
2720196
7118f9e
2114e35
 
 
 
 
 
 
 
7118f9e
2114e35
 
2720196
2114e35
 
 
2720196
2114e35
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2720196
2114e35
 
2720196
 
2114e35
2720196
2114e35
 
 
 
 
 
 
 
2720196
 
 
2114e35
 
 
 
 
 
2720196
 
2114e35
 
 
52f0f4a
2114e35
 
2720196
 
2114e35
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2720196
2114e35
 
 
 
 
 
2720196
2114e35
2720196
 
7f3c58e
2114e35
 
2720196
2114e35
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2720196
2114e35
2720196
 
2114e35
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
import gradio as gr
from pydub import AudioSegment
import json
import uuid
import edge_tts
import asyncio
import aiofiles
import os
import time
import mimetypes
import torch
import re
from typing import List, Dict
from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig

# Constants
MAX_FILE_SIZE_MB = 20
MAX_FILE_SIZE_BYTES = MAX_FILE_SIZE_MB * 1024 * 1024

MODEL_ID = "unsloth/gemma-3-1b-pt"

# Initialize model with proper error handling
try:
    tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
    if tokenizer.pad_token is None:
        tokenizer.pad_token = tokenizer.eos_token
    
    model = AutoModelForCausalLM.from_pretrained(
        MODEL_ID,
        torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
        device_map="auto",
        trust_remote_code=True
    ).eval()
    
    # Configure generation parameters
    generation_config = GenerationConfig(
        max_new_tokens=1024,
        temperature=0.7,
        top_p=0.9,
        do_sample=True,
        pad_token_id=tokenizer.pad_token_id,
        eos_token_id=tokenizer.eos_token_id,
    )
    
    print(f"Model loaded successfully on device: {model.device}")
    
except Exception as e:
    print(f"Model initialization error: {e}")
    model = None
    tokenizer = None
    generation_config = None

class PodcastGenerator:
    def __init__(self):
        self.model = model
        self.tokenizer = tokenizer
        self.generation_config = generation_config

    def extract_json_from_text(self, text: str) -> Dict:
        """Extract JSON from model output using regex patterns"""
        # Remove the input prompt from the output
        # Look for JSON-like structures
        json_patterns = [
            r'\{[^{}]*"topic"[^{}]*"podcast"[^{}]*\[.*?\]\s*\}',
            r'\{.*?"topic".*?"podcast".*?\[.*?\].*?\}',
        ]
        
        for pattern in json_patterns:
            matches = re.findall(pattern, text, re.DOTALL | re.IGNORECASE)
            for match in matches:
                try:
                    # Clean up the match
                    cleaned_match = match.strip()
                    return json.loads(cleaned_match)
                except json.JSONDecodeError:
                    continue
        
        # If no valid JSON found, create a fallback structure
        return self.create_fallback_podcast(text)

    def create_fallback_podcast(self, text: str) -> Dict:
        """Create a basic podcast structure when JSON parsing fails"""
        # Extract meaningful sentences from the text
        sentences = [s.strip() for s in text.split('.') if len(s.strip()) > 10]
        
        if not sentences:
            sentences = ["Let's discuss this interesting topic.", "That's a great point to consider."]
        
        podcast_lines = []
        for i, sentence in enumerate(sentences[:10]):  # Limit to 10 exchanges
            speaker = (i % 2) + 1
            podcast_lines.append({
                "speaker": speaker,
                "line": sentence + "." if not sentence.endswith('.') else sentence
            })
        
        return {
            "topic": "Generated Discussion",
            "podcast": podcast_lines
        }

    async def generate_script(self, prompt: str, language: str, file_obj=None, progress=None) -> Dict:
        if not self.model or not self.tokenizer:
            raise Exception("Model not properly initialized. Please check model loading.")

        example_json = {
            "topic": "AGI",
            "podcast": [
                {"speaker": 1, "line": "So, AGI, huh? Seems like everyone's talking about it these days."},
                {"speaker": 2, "line": "Yeah, it's definitely having a moment, isn't it?"},
                {"speaker": 1, "line": "It really is. What got you hooked on this topic?"},
                {"speaker": 2, "line": "The potential implications are fascinating and concerning at the same time."}
            ]
        }

        if language == "Auto Detect":
            language_instruction = "Use the same language as the input text"
        else:
            language_instruction = f"Generate the podcast in {language} language"

        # Simplified, more direct prompt
        system_prompt = f"""Generate a podcast script as valid JSON. {language_instruction}.

Requirements:
- Exactly 2 speakers (speaker 1 and 2)
- Natural, engaging conversation
- JSON format only

Example format:
{json.dumps(example_json, indent=2)}

Input topic: {prompt}

Generate JSON:"""

        try:
            if progress:
                progress(0.3, "Generating podcast script...")

            # Tokenize with proper attention mask
            inputs = self.tokenizer(
                system_prompt, 
                return_tensors="pt", 
                padding=True, 
                truncation=True,
                max_length=2048
            )
            inputs = {k: v.to(self.model.device) for k, v in inputs.items()}

            # Generate with timeout
            with torch.no_grad():
                output = self.model.generate(
                    **inputs,
                    generation_config=self.generation_config,
                    pad_token_id=self.tokenizer.pad_token_id,
                )

            # Decode only the new tokens
            generated_text = self.tokenizer.decode(
                output[0][inputs['input_ids'].shape[1]:], 
                skip_special_tokens=True
            )

            print(f"Generated text: {generated_text[:500]}...")

            if progress:
                progress(0.4, "Processing generated script...")

            # Extract JSON from the generated text
            result = self.extract_json_from_text(generated_text)
            
            if progress:
                progress(0.5, "Script generated successfully!")
            
            return result

        except Exception as e:
            print(f"Generation error: {e}")
            # Return fallback podcast
            return {
                "topic": prompt or "Discussion",
                "podcast": [
                    {"speaker": 1, "line": f"Welcome to our discussion about {prompt or 'this topic'}."},
                    {"speaker": 2, "line": "Thanks for having me. This is indeed an interesting subject."},
                    {"speaker": 1, "line": "Let's dive into the key points and explore different perspectives."},
                    {"speaker": 2, "line": "Absolutely. There's a lot to unpack here."},
                    {"speaker": 1, "line": "What aspects do you find most compelling?"},
                    {"speaker": 2, "line": "The implications and potential applications are fascinating."},
                    {"speaker": 1, "line": "That's a great point. Thanks for the insightful discussion."},
                    {"speaker": 2, "line": "Thank you. This has been a valuable conversation."}
                ]
            }

    async def tts_generate(self, text: str, speaker: int, speaker1: str, speaker2: str) -> str:
        """Generate TTS audio with improved error handling"""
        voice = speaker1 if speaker == 1 else speaker2
        speech = edge_tts.Communicate(text, voice)
        
        temp_filename = f"temp_audio_{uuid.uuid4()}.wav"
        max_retries = 3
        
        for attempt in range(max_retries):
            try:
                await asyncio.wait_for(speech.save(temp_filename), timeout=30)
                if os.path.exists(temp_filename) and os.path.getsize(temp_filename) > 0:
                    return temp_filename
                else:
                    raise Exception("Generated audio file is empty")
            except asyncio.TimeoutError:
                if os.path.exists(temp_filename):
                    os.remove(temp_filename)
                if attempt == max_retries - 1:
                    raise Exception("TTS generation timed out after multiple attempts")
                await asyncio.sleep(1)  # Brief delay before retry
            except Exception as e:
                if os.path.exists(temp_filename):
                    os.remove(temp_filename)
                if attempt == max_retries - 1:
                    raise Exception(f"TTS generation failed: {str(e)}")
                await asyncio.sleep(1)

    async def combine_audio_files(self, audio_files: List[str], progress=None) -> str:
        """Combine audio files with silence padding"""
        if progress:
            progress(0.9, "Combining audio files...")
            
        try:
            combined_audio = AudioSegment.empty()
            silence_padding = AudioSegment.silent(duration=500)  # 500ms silence
            
            for i, audio_file in enumerate(audio_files):
                try:
                    audio_segment = AudioSegment.from_file(audio_file)
                    combined_audio += audio_segment
                    
                    # Add silence between speakers (except for the last file)
                    if i < len(audio_files) - 1:
                        combined_audio += silence_padding
                        
                except Exception as e:
                    print(f"Warning: Could not process audio file {audio_file}: {e}")
                finally:
                    # Clean up temporary file
                    if os.path.exists(audio_file):
                        os.remove(audio_file)

            if len(combined_audio) == 0:
                raise Exception("No audio content generated")

            output_filename = f"podcast_output_{uuid.uuid4()}.wav"
            combined_audio.export(output_filename, format="wav")
            
            if progress:
                progress(1.0, "Podcast generated successfully!")
                
            return output_filename
            
        except Exception as e:
            # Clean up any remaining temp files
            for audio_file in audio_files:
                if os.path.exists(audio_file):
                    os.remove(audio_file)
            raise Exception(f"Audio combination failed: {str(e)}")

    async def generate_podcast(self, input_text: str, language: str, speaker1: str, speaker2: str, file_obj=None, progress=None) -> str:
        """Main podcast generation pipeline with improved error handling"""
        try:
            if progress:
                progress(0.1, "Starting podcast generation...")

            # Generate script
            podcast_json = await self.generate_script(input_text, language, file_obj, progress)
            
            if not podcast_json.get('podcast'):
                raise Exception("No podcast content generated")

            if progress:
                progress(0.5, "Converting text to speech...")

            # Generate TTS with sequential processing to avoid overload
            audio_files = []
            total_lines = len(podcast_json['podcast'])
            
            for i, item in enumerate(podcast_json['podcast']):
                try:
                    audio_file = await self.tts_generate(
                        item['line'], 
                        item['speaker'], 
                        speaker1, 
                        speaker2
                    )
                    audio_files.append(audio_file)
                    
                    # Update progress
                    if progress:
                        current_progress = 0.5 + (0.4 * (i + 1) / total_lines)
                        progress(current_progress, f"Generated speech {i + 1}/{total_lines}")
                        
                except Exception as e:
                    print(f"TTS error for line {i}: {e}")
                    # Continue with remaining lines
                    continue

            if not audio_files:
                raise Exception("No audio files generated successfully")

            # Combine audio files
            combined_audio = await self.combine_audio_files(audio_files, progress)
            return combined_audio

        except Exception as e:
            raise Exception(f"Podcast generation failed: {str(e)}")

# Voice mapping
VOICE_MAPPING = {
    "Andrew - English (United States)": "en-US-AndrewMultilingualNeural",
    "Ava - English (United States)": "en-US-AvaMultilingualNeural",
    "Brian - English (United States)": "en-US-BrianMultilingualNeural",
    "Emma - English (United States)": "en-US-EmmaMultilingualNeural",
    "Florian - German (Germany)": "de-DE-FlorianMultilingualNeural",
    "Seraphina - German (Germany)": "de-DE-SeraphinaMultilingualNeural",
    "Remy - French (France)": "fr-FR-RemyMultilingualNeural",
    "Vivienne - French (France)": "fr-FR-VivienneMultilingualNeural"
}

async def process_input(input_text: str, input_file, language: str, speaker1: str, speaker2: str, progress=None) -> str:
    """Process input and generate podcast"""
    start_time = time.time()

    try:
        if progress:
            progress(0.05, "Processing input...")

        # Map speaker names to voice IDs
        speaker1_voice = VOICE_MAPPING.get(speaker1, "en-US-AndrewMultilingualNeural")
        speaker2_voice = VOICE_MAPPING.get(speaker2, "en-US-AvaMultilingualNeural")

        # Validate input
        if not input_text or input_text.strip() == "":
            if input_file is None:
                raise Exception("Please provide either text input or upload a file")
            # TODO: Add file processing logic here if needed

        podcast_generator = PodcastGenerator()
        result = await podcast_generator.generate_podcast(
            input_text, language, speaker1_voice, speaker2_voice, input_file, progress
        )

        end_time = time.time()
        print(f"Total generation time: {end_time - start_time:.2f} seconds")
        return result

    except Exception as e:
        error_msg = str(e)
        print(f"Processing error: {error_msg}")
        raise Exception(f"Generation failed: {error_msg}")

def generate_podcast_gradio(input_text, input_file, language, speaker1, speaker2):
    """Gradio interface function with proper error handling"""
    try:
        # Validate inputs
        if not input_text and input_file is None:
            return None
            
        if input_text and len(input_text.strip()) == 0:
            input_text = None

        # Create a simple progress tracker
        progress_history = []
        
        def progress_callback(value, text):
            progress_history.append(f"{value:.1%}: {text}")
            print(f"Progress: {value:.1%} - {text}")

        # Run the async function
        loop = asyncio.new_event_loop()
        asyncio.set_event_loop(loop)
        try:
            result = loop.run_until_complete(
                process_input(input_text, input_file, language, speaker1, speaker2, progress_callback)
            )
            return result
        finally:
            loop.close()
            
    except Exception as e:
        print(f"Gradio function error: {e}")
        raise gr.Error(f"Failed to generate podcast: {str(e)}")

def create_interface():
    """Create the Gradio interface with proper component configuration"""
    language_options = [
        "Auto Detect", "English", "German", "French", "Spanish", "Italian", 
        "Portuguese", "Dutch", "Russian", "Chinese", "Japanese", "Korean"
    ]
    
    voice_options = list(VOICE_MAPPING.keys())
    
    with gr.Blocks(
        title="PodcastGen 2🎙️",
        theme=gr.themes.Soft(),
        css=".gradio-container {max-width: 1200px; margin: auto;}"
    ) as demo:
        
        gr.Markdown("# 🎙️ PodcastGen 2")
        gr.Markdown("Generate professional 2-speaker podcasts from text input!")
        
        with gr.Row():
            with gr.Column(scale=2):
                input_text = gr.Textbox(
                    label="Input Text",
                    lines=8,
                    placeholder="Enter your topic or text for podcast generation...",
                    info="Describe what you want the podcast to discuss"
                )
            
            with gr.Column(scale=1):
                input_file = gr.File(
                    label="Upload File (Optional)",
                    file_types=[".pdf", ".txt"],
                    type="filepath",
                    info=f"Max size: {MAX_FILE_SIZE_MB}MB"
                )
        
        with gr.Row():
            language = gr.Dropdown(
                label="Language",
                choices=language_options,
                value="Auto Detect",
                info="Select output language"
            )
            
            speaker1 = gr.Dropdown(
                label="Speaker 1 Voice",
                choices=voice_options,
                value="Andrew - English (United States)"
            )
            
            speaker2 = gr.Dropdown(
                label="Speaker 2 Voice",
                choices=voice_options,
                value="Ava - English (United States)"
            )
        
        generate_btn = gr.Button(
            "🎙️ Generate Podcast",
            variant="primary",
            size="lg"
        )
        
        output_audio = gr.Audio(
            label="Generated Podcast",
            type="filepath",
            format="wav",
            show_download_button=True
        )
        
        # Connect the interface
        generate_btn.click(
            fn=generate_podcast_gradio,
            inputs=[input_text, input_file, language, speaker1, speaker2],
            outputs=[output_audio],
            show_progress=True
        )
        
        # Add usage instructions
        with gr.Accordion("Usage Instructions", open=False):
            gr.Markdown("""
            ### How to use:
            1. **Input**: Enter your topic or text in the text box, or upload a PDF/TXT file
            2. **Language**: Choose the output language (Auto Detect recommended)
            3. **Voices**: Select different voices for Speaker 1 and Speaker 2
            4. **Generate**: Click the button and wait for processing
            
            ### Tips:
            - Provide clear, specific topics for better results
            - The AI will create a natural conversation between two speakers
            - Generation may take 1-3 minutes depending on text length
            """)
    
    return demo

if __name__ == "__main__":
    demo = create_interface()
    demo.launch(
        server_name="0.0.0.0",
        server_port=7860,
        show_error=True,
        share=False
    )