How to Down-Pitch A Song Using Python

If you’ve ever wanted to change the pitch of a song without altering its speed, this blog post is for you. Pitch-shifting is a common task for musicians, DJs, and audio engineers. In this tutorial, we will explore how to down-pitch a song using Python and the pydub library and apply this process to multiple songs in a folder automatically.

Why Pitch Shift?

In music, pitch-shifting means changing the pitch of a song (raising or lowering it) without speeding it up or slowing it down. This can be useful for:

• Matching the key of a song to another track

• Transposing songs for instruments tuned to a different key

• Creating remixes or mashups

Tools You Will Need

We will use the Python library pydub to manipulate audio files. You can install it using pip:

pip install pydub

Additionally, pydub requires ffmpeg to handle audio files like MP3. You can install ffmpeg via the terminal:

sudo apt install ffmpeg

Step-by-Step Guide to Pitch Shifting

Now let’s dive into the Python script that automates pitch-shifting for multiple songs in a folder. The script loops through the files in a songs folder, down-pitches them by a half-step (semitone = -1), and saves the new files to an output folder.

The Code

import os
from pydub import AudioSegment

# Function to shift pitch down
def pitch_shift(audio, semitones):
    # Adjust sample rate to shift pitch
    new_sample_rate = int(audio.frame_rate * (2.0 ** (semitones / 12.0)))
    return audio._spawn(audio.raw_data, overrides={'frame_rate': new_sample_rate}).set_frame_rate(audio.frame_rate)

# Input and output folders
input_folder = './songs'
output_folder = './output'

# Ensure the output folder exists
os.makedirs(output_folder, exist_ok=True)

# Loop through all files in the songs folder
for filename in os.listdir(input_folder):
    # Check if the file is an audio file (e.g., mp3 or wav)
    if filename.endswith(".mp3") or filename.endswith(".wav"):
        # Construct the full file path
        input_path = os.path.join(input_folder, filename)
        output_path = os.path.join(output_folder, filename)

        # Load the audio file
        audio = AudioSegment.from_file(input_path)

        # Shift pitch down by a half-step (semitone = -1)
        shifted_audio = pitch_shift(audio, -1)

        # Export the pitch-shifted audio to the output folder
        shifted_audio.export(output_path, format="mp3")
        print(f"Processed and saved: {output_path}")

Explanation

1. Importing Libraries:
   We import os to work with file directories and AudioSegment from pydub to manipulate audio files.

2. Pitch-Shift Function:
   The pitch_shift function adjusts the sample rate of the audio. When we change the sample rate, the pitch changes. In this case, we calculate the new sample rate to shift the pitch down by one semitone using the formula:
   
   new_sample_rate = int(audio.frame_rate * (2.0 ** (semitones / 12.0)))
   

3. Input and Output Folders:
   We define the folders where we will read the audio files and save the pitch-shifted versions. If the output folder doesn't exist, it will be created.

4. Loop Through Songs:
   Using os.listdir(), we loop through each file in the songs folder. The script checks if the file is an audio file (.mp3 or .wav) before processing it. For each file:

   • It loads the audio.
   • The pitch_shift function is applied, lowering the pitch by a half-step.
   • The pitch-shifted audio is exported to the output folder.

5. Export and Feedback:
   Once the processing is done, the pitch-shifted song is saved in the output folder, and a confirmation message is printed.

Running the Script

Make sure you have your audio files in the songs folder and then run the script:

python -m pitch_down

The pitch-shifted files will be saved in the output folder.

Customization

You can easily modify this script to:

• Pitch the audio up by passing a positive value (e.g., pitch_shift(audio, 1) for a half-step up).

• Process different file formats by adding other extensions like .ogg or .flac to the conditional check.

• Shift by a different number of semitones by adjusting the semitones argument.

Conclusion

This script is a simple yet powerful way to pitch-shift multiple audio files using Python. With pydub and ffmpeg, you can manipulate audio files in bulk, making tasks like pitch correction or audio preparation easier for musicians, producers, or anyone working with audio.

Feel free to experiment with this script and see how you can adapt it to your needs. Happy coding!

Comments

[Scott Reno]

Wow this is really cool!

[Vicente G. Reyes]

Thanks!!

[Sam J.]

Can you explain the new sample formula a bit more in detail, like the magic numbers?

[Vicente G. Reyes]

What this does:

This formula changes the pitch of an audio file by adjusting its sample rate based on the number of semitones you want to shift up or down.
Key Terms:

1. audio.frame_rate: This is the original sample rate of the audio file. It defines how many samples per second the audio was originally recorded at.

2. semitones: A semitone is the smallest musical interval (the distance between two adjacent keys on a piano). Positive values shift the pitch up, and negative values shift it down.

3. 2.0 ** (semitones / 12.0): This is the mathematical formula that determines the pitch shift. The number 12 represents the number of semitones in an octave. So:
   If semitones = 12, you shift the pitch up by 1 full octave.
   If semitones = -12, you shift the pitch down by 1 full octave.
   If semitones = 1, you shift the pitch up by 1 semitone (like moving one key up on a piano).

4. new_sample_rate: This is the sample rate of the audio after the pitch shift.

How it works:

• Higher semitones (positive values) will make the pitch higher by increasing the sample rate.
• Lower semitones (negative values) will make the pitch lower by decreasing the sample rate.

Finally, int() is used to ensure the result is a whole number, as the sample rate must be an integer.
Example:

• If audio.frame_rate = 44100 Hz (a common sample rate) and you want to shift up by 1 semitone:

  python

new_sample_rate = int(44100 * (2.0 ** (1 / 12.0)))

• This increases the sample rate slightly, resulting in a pitch shift upwards by 1 semitone.