import json
import os
import random

def seconds_to_mmss(seconds):
    minutes = int(seconds // 60)
    seconds = int(seconds % 60)
    return f"{minutes:02d}:{seconds:02d}"

# Templates for overlap descriptions
OVERLAP_TEMPLATES = [
    "An overlap where multiple speakers talk simultaneously for more than 3 seconds starts at {start} and ends at {end}.",
    "The overlap starts at {start} and ends at {end}.",
    "Multiple speakers talk simultaneously from {start} to {end}, an overlap lasting over three seconds.",
    "A conversation overlap occurs between {start} and {end}, with multiple people speaking at once.",
    "There is a significant overlap in the conversation from {start} to {end}, where speakers talk simultaneously.",
    "During the period {start}-{end}, multiple participants speak at the same time.",
    "A 3+ second overlap is detected between {start} and {end}, with concurrent speakers.",
    "The conversation features overlapping speech from {start} to {end}.",
    "Several speakers talk over each other between {start} and {end}.",
    "An overlapping segment is identified from {start} to {end}, lasting more than 3 seconds."
]

# Templates for no overlap case
NO_OVERLAP_TEMPLATES = [
    "No significant overlaps found in this conversation.",
    "The conversation shows no overlapping speech segments longer than 3 seconds.",
    "No instances of multiple speakers talking simultaneously for more than 3 seconds were detected.",
    "The dialogue proceeds without any significant overlaps between speakers.",
    "No overlapping segments exceeding 3 seconds were identified in this conversation.",
    "The speakers maintain clear turn-taking without significant overlaps.",
    "This conversation features no substantial overlapping speech periods.",
    "No overlapping speech segments of 3 seconds or longer were found.",
    "The conversation flows smoothly without any significant speaker overlaps.",
    "All speakers maintain clear speaking turns without substantial overlap."
]

# Templates for correct overlap case
CORRECT_OVERLAP_TEMPLATES = [
    "The conversation contains overlapping speech segments longer than 3 seconds.",  
    "The dialogue features instances where speakers overlap for more than 3 seconds.",  
    "The recording includes multiple overlapping exchanges exceeding 3 seconds.",  
    "There are clear overlaps in speech lasting beyond 3 seconds.",  
    "Speaker interruptions or overlaps exceed 3 seconds in duration.",  
    "Clear evidence shows speech overlaps extending beyond 3 seconds in duration.",
    "Speaker overlap durations consistently breach the 3-second threshold.",
    "Recorded overlaps between speakers routinely last longer than 3 seconds.",
    "The interaction contains several instances where voices overlap for over 3 seconds."
]
file = "silence"
def process_overlap_segments():
    # Read the overlap_5s_716.json file
    with open(f'{file}.json', 'r', encoding='utf-8') as f:
        overlap_data = json.load(f)

    # List to store results for all conversations
    results = []

    # Process each conversation
    for conversation_id, conversation in overlap_data.items():
        segments = conversation.get('segments', [])
        overlap_periods = []
        audio_path = conversation.get('stereo_audio', [])
        # Find overlaps > 5s between segments
        for i in range(len(segments) - 1):
            current = segments[i]
            next_segment = segments[i + 1]
            
            # Calculate overlap
            overlap_start = max(current['start_time'], next_segment['start_time'])
            overlap_end = min(current['end_time'], next_segment['end_time'])
            overlap_duration = overlap_end - overlap_start
            
            # If overlap is more than 5 seconds
            if overlap_duration >= 3:
                overlap_periods.append(f"{seconds_to_mmss(overlap_start)}-{seconds_to_mmss(overlap_end)}")

        # Create result entry with random template
        if overlap_periods:
            # For each overlap period, select a random template
            overlap_descriptions = []
            for period in overlap_periods:
                start, end = period.split('-')
                template = random.choice(OVERLAP_TEMPLATES)
                overlap_descriptions.append(template.format(start=start, end=end))
            model_output = " ".join(overlap_descriptions)
        else:
            model_output = random.choice(NO_OVERLAP_TEMPLATES)

        result = {
            "key": conversation_id,
            "audio_url": audio_path,
            "model_output": model_output
        }
        results.append(result)

    # Save the results to a JSON file
    output_file = f'{file}_overlapgap.json'
    with open(output_file, 'w', encoding='utf-8') as f:
        json.dump(results, f, indent=2, ensure_ascii=False)
    
    print(f"Processed {len(results)} conversations")
    print(f"Results written to {output_file}")

if __name__ == "__main__":
    process_overlap_segments()