Data Makes Music – Spotify writes Udio Pompts

mkdir promptmaker
    cd promptmaker
    
    # setup to isolate the installation 
    virtualenv env
    
    # activate environment
    source env/bin/activate
    
    # install necessary libraries
    pip3 install spotipy openai

    export SPOTIPY_CLIENT_ID='your-spotify-client-id'
    export SPOTIPY_CLIENT_SECRET='your-spotify-client-secret'
    export SPOTIPY_REDIRECT_URI='your-app-redirect-url'
    # Again, make sure you have the forward slash at the end
    
    # If you want to analyze the song titles 
    # and have access to openAI API, enter the key
    export OPENAI_API_KEY='your-open-ai-api-key'

import spotipy
    from spotipy.oauth2 import SpotifyOAuth
    from math import floor
    from collections import Counter
    from openai import OpenAI
    
    number_of_top_tracks = 10 # keep it between 3 and 50
    scope = 'user-top-read' # submitted to spotify for the authentication process
    
    sp = spotipy.Spotify(auth_manager=SpotifyOAuth(scope=scope))
    results = sp.current_user_top_tracks(limit=number_of_top_tracks,
                                            offset=0, 
                                            time_range='long_term')
    
    # helper function for prompt building
    def prompt_from_range(input_list, input_val):
        """returns the value matching the input on a range 0 to 1"""
        return_item = input_list[floor(input_val * len(input_list))]
        return f'{return_item}, '
    
    # init variables
    counter = 0
    acousticness_sum = 0
    danceability_sum = 0
    energy_sum = 0
    tempo_sum = 0
    valence_sum = 0
    instrumentalness_sum = 0
    liveness_sum = 0
    mode_sum = 0
    genre_list = []
    title_list = []

for i, track in enumerate(results['items']):
        print('---')
        title = track['name']
        artist_name = track['artists'][0]['name']
        print(f'{i} {title} – {artist_name}')
    
        track_id = track['id']
        audio_features = sp.audio_features(track_id)
    
        speechiness = audio_features[0]['speechiness']
        if speechiness > 0.5:
            # ignore spoken content 
            print(f'speechiness: {speechiness} – ignoring as spoken content')
        else:
            counter += 1
            title_list.append(title)
    
            acousticness = audio_features[0]['acousticness']
            acousticness_sum += acousticness
    
            danceability = audio_features[0]['danceability']
            danceability_sum += danceability
            
            energy = audio_features[0]['energy']
            energy_sum += energy
    
            tempo = audio_features[0]['tempo']
            tempo_sum += tempo
    
            valence = audio_features[0]['valence']
            valence_sum += valence
    
            instrumentalness = audio_features[0]['instrumentalness']
            instrumentalness_sum += instrumentalness
            
            liveness = audio_features[0]['liveness']
            liveness_sum += liveness
    
            mode = audio_features[0]['mode']
            mode_sum += mode

            artist_id = track['artists'][0]['id']
            artist_info = sp.artist(artist_id)
            genres = artist_info.get('genres')
            print(f'Genres: {genres}')
            genre_list += genres #build a genre list

try:
        # get the most prominent genres
        genre_counts = Counter(genre_list) # creates a list of tuples  
        common_genres = genre_counts.most_common(3) # sort and limit 
        keys = [genre for genre, count in common_genres] 

        # find the longest genre description which is often the funniest
        max_length = max(len(genre) for genre in genre_list)
        peculiar_genre = [genre for genre in genre_list 
                            if len(genre) == max_length]
        peculiar_genre = peculiar_genre[0]

        genre_prompt = f'A {keys[0]} song with {keys[1]} and {keys[2]} elements.'
        genre_prompt += f'Spice it up with some {peculiar_genre}, '
    
        prompt = genre_prompt + prompt

    except Exception as e:
        # sometimes no genres are defined and the handling fails. 
        print(f'Error {e} while handling genres. Proceeding without genres.')

try: 
        client = OpenAI()
        # chose the model: "gpt-3.5-turbo" (cheaper) or "gpt-4" (costier, better).
        model = "gpt-4"

        system_prompt = """You are a record company's portfolio 
            manager who analyzes 
            societal, behavioural, emotional topics in song titles.
            He considers several titles and transforms the most 
            striking meta-topic into brief for a lyricist that starts: 
            Write a song about [meta-topic]."""

        # Extrapolate the topic from your top 10 tracks 
        user_prompt = f"""Please have a look at these song titles {title_list[:10]} 
                            and turn them into a short single-sentence brief."""

        completion = client.chat.completions.create(
        model=model,
        messages=[  {"role": "system", "content": system_prompt},
                    {"role": "user", "content": user_prompt}])
        lyrics_prompt = completion.choices[0].message.content + " "

    except Exception as e:
        print(f'Error {e} while fetching lyrics. Proceeding with generic prompt')
        lyrics_prompt = f'A song about \'{title_list[0]}\' '
        lyrics_prompt += f'from someone who adores \'{title_list[1]}\' '
        lyrics_prompt += f'while thinking \'{title_list[2]}\'.'

    prompt = lyrics_prompt + prompt

    ### That’s all we need for the prompt. Print it to the Terminal:

    print('\n\nHere is the prompt generated from your data.')
    print('Copy and paste it into your song generation tool.')
    print('---')
    print(prompt)
    print('---')