user_variables.py

""" USER CHOICES """

############# MANUAL SECTION 1 : Variables ####################

""" Choose if the program should run with (True) or without (False) a log file.
- Set to False to run this script without a log file.
- True will require log file instanciation for the program to proceed (if an error prevents this, the program will exit)."""
use_log_file = True # True or False only

""" Choose whether log files should be saved in a separate folder.
- If yes, specify the directory path (relative or absolute)
- Supply an empty string "" to use the program directory, or a directory path i.e. "output/".
- Path cannot be None.
- Directory will be created if doesn't exist.
"""
path_to_logs = "logs/" # supply empty string "" for the program directory, or a directory path. Do not use None.


""" Specify directory path containing audio files to be transcribed.
- For relative paths, supply empty string "" to use the program directory, or a directory path i.e. "output/". 
- Path cannot be None.
- Files in subfolders will NOT be enumerated or processed.
- If the directory does not exist, program will exit with an error message."""
path_to_audio = "batch/" 


"""Specify output folder (relative) or full path (absolute) where transcripts will be saved.
For relative paths, supply empty string "" to use the program directory, or a directory path i.e. "output/".
- Path cannot be None.
- Directory will be created if doesn't exist."""
path_for_output = "output/"

"""Choose whether processed audio files should be moved to a new folder after transcription.
- Directory can be the same as path_for_output, or different.
- For relative paths, supply empty string "" to use the program directory, or a directory path i.e. "output/". 
- Path cannot be None.
- Directory will be created if doesn't exist. """
move_processed = True # True or False only
path_for_processed = "output/" 


"""Specify target audio file format.
- only one file type at a time can be processed
- include the . and make sure the case matches the actual file extension:
".wav" will not match if the files are ".WAV" """
audio_format = ".mp4" # Note: 


"""Choose the Whisper transcription model"""
# Whisper Model Options (tiny & base 1GB VRAM, small 2GB, medium 5GB, large 10GB). Non .en are multilingual. en's are best for English.

model_options = {
    "Tiny_English": {"name": "tiny.en", "speed_x": 32, "alt_name": "tiny_en"},
    "Base_English": {"name": "base.en", "speed_x": 16, "alt_name": "base_en"},
    "Small_English": {"name": "small.en", "speed_x": 6, "alt_name": "small_en"},
    "Medium_English": {"name": "medium.en", "speed_x": 2, "alt_name": "medium_en"},
    "Tiny_Multilingual": {"name": "tiny", "speed_x": 32, "alt_name": "tiny_multi"},
    "Base_Multilingual": {"name": "base", "speed_x": 16, "alt_name": "base_multi"},
    "Small_Multilingual": {"name": "small", "speed_x": 6, "alt_name": "small_multi"},
    "Medium_Multilingual": {"name": "medium", "speed_x": 2, "alt_name": "medium_multi"},
    "Large_Multilingual": {"name": "large", "speed_x": 1, "alt_name": "large_multi"}
    }

""" 
- Supply the dictionary key ("Tiny_English"), NOT the model name ("tiny.en")
- See print statements (each file) / log entry (file batch) produced by process_time_estimator() for processing time estimations."""
model_key = "Medium_English" # spelling must match exactly. 


""" Choose word interval for line wrapping and to insert line-numbers into the final transcript.
- Whisper returns transcripts which are one long string of text with no linebreaks or speaker labels. Therefore:
- Specify the interval of words at which to insert a newline into transcript, or
- Enter 0 to return the raw transcript with no line wrapping and no line numbers
- If an invalid interval is entered, 0 will be substituted.
- ! Important note: choosing in interval will not just wrap the lines, but will insert a line number at the start of each line in the format XX: (i.e. 12:). If you don't want line numbers, set the word interval to 0."""
word_interval = 10  # approx 9 - 12 is English average


"""Choose a delimiter for transcript. 
- Supply an empty string "" to skip inserting a delimiter."""
delimiter = "---"


############# MANUAL SECTION 2 : POPULATE TRANSCRIPT HEADER ####################

"""OPTION 1: Populate header fields as a group.

Fill in any string fields which are the same for all audio files in the batch."""

# If any fields are not required, do NOT remove them here: go to create_header() in whisper_wrapper.py,commenting out the line beginning `f"{field_name}` which you want to omit.

audio_info_batch = [
    {
        "participants" : [
        {"name": "", "role": ""}, # Host, Speaker, Interviewer
        {"name": "", "role": ""},
        ]
    },
    {
        "audio_content" : [
        {"type" : ""}, # "talk", "interview", "QandA"
        {"topic" : ""}, # e.g. "Health and Safety", "Business", "Economics"
        {"series" : ""}, # e.g. "BBC The Bottom Line"
        {"format" : ""}, # e.g. "Podcast", "YouTube"
        {"date" : ""} # most content will be date unique, and so this field is likely to need to be commented out in create_header() in whisper_wrapper.py
        ] 
    },
    {
        "transcript_type" : [
        {"producer" : "whisper"},
        {"model" : model_options[model_key]["name"]}
        ]
    },
]

"""OPTION 2: Populate header fields which vary per audio file.

This option is laborious and NOT recommended: see the notes in README.md. By default, the create_header() function points towards the `audio_info_batch` dictionary above, so you will have to change the function to point towards the relevant key in the `audio_file_info` dictionary below. The commented code provides the alternative reference to each lines.
"""

audio_file_info = [
    # Individualised nfo for file 1
    {
    "index" : 1, # {audio_file_info[index-1]['index']}
    "date" : "2024-02-08", # {audio_file_info[index-1]['date']}
    
    # NB: only need series_num and episode_num if the audio file name does not contain this information: S06 E07 can be extracted by extract_series_episode()
    "series_num" : "S6", # {audio_file_info[index-1]['series_num']}
    "episode_num" : "E7", # {audio_file_info[index-1]['episode_num']}
    
    "speakers" : [
        {"name": "Mr Presenter", "role": "Host"}, # Host, Speaker, Interviewer
        {"name": "Mr A", "role": "Interviewee"},
        {"name": "Mrs B", "role": "Interviewee"},
       ]
    },
    # Individualised info for file 2
    {
    "index" : 2, # {audio_file_info[index-1]['index']}
    "date" : "2024-02-15", # {audio_file_info[index-1]['date']}
    
    # NB: only need series_num and episode_num if the file name does not contain this information: S06 E07 can be extracted by extract_series_episode()
    "series_num" : "S6", # {audio_file_info[index-1]['series_num']}
    "episode_num" : "E7", # {audio_file_info[index-1]['episode_num']}
    
    "speakers" : [
        {"name": "Dr Bob", "role": "Host"}, # Host, Speaker, Interviewer
        {"name": "Spongebob", "role": "Interviewee"},
        {"name": "Catty McCat", "role": "Guest"},
       ]
    } # Add as many dictionaries as there are files.
]