find_steps.py

from sentence_transformers import SentenceTransformer, util

import json
from datetime import datetime
# Function to read and parse the JSON file
def find_next_step(data, current_context):
    try:
        # Read the JSON file
        # with open(file_path, 'r') as file:
        #     data = json.load(file)
        
        # Initialize counters
        total_videos = len(data)
        total_scenes = 0
        max_score=0.0
        step={
            "path" :"",
            "start":"",
            "end": "",
            }
        # Iterate through each video entry
        print(f"Processing {total_videos} videos...\n")
        for index, video in enumerate(data, 1):
            video_type = video.get('type', 'Unknown')
            video_path = video.get('path', 'No path provided')
            scenes = video.get('scenes', [])
            scene_count = len(scenes)
            total_scenes += scene_count
            

            # Iterate through each scene in the video
            for scene_index, scene in enumerate(scenes, 1):
                timestamp = scene.get('timestamp', 'No timestamp')
                description = scene.get('description', 'No description')

                time_obj = datetime.strptime(timestamp, "%M:%S.%f")
                start = time_obj.minute * 60 + time_obj.second + time_obj.microsecond / 1_000_000               
                score=compare_phrases(description,current_context)

                if score > max_score:
                    if scene_index+1<len(scenes):
                        next_scene=scenes[scene_index+1]
                        next_timestamp=next_scene.get('timestamp', 'No timestamp')
                        time_obj = datetime.strptime(next_timestamp, "%M:%S.%f")

                        stop = time_obj.minute * 60 + time_obj.second + time_obj.microsecond / 1_000_000  
                        step={
                        "path" :video_path,
                        "start":start,
                        "end": stop,
                        "description":description
                        }
                    else:
                        step={
                        "path" :video_path,
                        "start":start,
                        "end": "",
                        "description":description
                        }

                    max_score=score

                if "Error" in description:
                    print("[Note: This scene contains an error]")
        
        # Print overall summary
        print(f"Summary:")
        print(f"  Total Videos: {total_videos}")
        print(f"  Total Scenes: {total_scenes}")
        
    except FileNotFoundError:
        print(f"Error: File '{file_path}' not found.")
    except json.JSONDecodeError:
        print("Error: Invalid JSON format.")
    except Exception as e:
        print(f"Error: An unexpected error occurred: {str(e)}")

    return step



def compare_phrases(phrase1,phrase2="Passing around a dj mixing a song"):
    model = SentenceTransformer('all-MiniLM-L6-v2',local_files_only=True)
    # Get embeddings
    embedding1 = model.encode(phrase1, convert_to_tensor=True)
    embedding2 = model.encode(phrase2, convert_to_tensor=True)

    # Compute cosine similarity
    similarity_score = util.cos_sim(embedding1, embedding2)[0][0]


    # print(f"Similarity score: {similarity_score:.4f}")
    # print(phrase1)
    return similarity_score



def find_all_steps(data, context):

    context = context.replace("\n", "").lower()
    all_actions=context.split(".")
    steps=[]
    for action in all_actions:
        step=find_next_step(data,action)
        steps.append(step)
    return steps


    


if __name__ == "__main__":
    file_path = '/Users/georgia.bucea/products/ShortsAI/all_files_metadata.json'  # Update this with the actual file path

    context = """I am walking around San Francisco and passing around the big white building. I am walking around a music band and DJ mixing a song. I am showing a demo on the phone, I am explaining the demo on the mobile phone"""
    # Read the JSON file
    with open(file_path, 'r') as file:
         data = json.load(file)

    steps=find_all_steps(data,context)
    print(steps)