Rock n Roll the LLMs

Can I generate a rock n roll using LLMs?
LLMs
music
prompt
sono
Author

Emoo

Published

April 16, 2024

from IPython.display import Audio

Here are some AI-powered music generation platforms:

  1. Amper Music: An AI music composition platform that allows users to create custom music tracks in minutes.
  2. AIVA: An AI-powered music generation platform that creates original music for films, ads, and video games.
  3. Jukedeck: A UK-based startup that uses AI to create music for videos, ads, and other media.
  4. Endel: An AI-powered music generation platform that creates personalized soundscapes for relaxation, focus, and sleep.
  5. Magix Music Maker: A music creation software that uses AI to help users create music tracks.
  6. Logic Pro X: A professional digital audio workstation (DAW) that uses AI-powered tools for music creation and editing.
  7. Flow Machines: A European research project that uses AI to generate music in various styles.
  8. Daddy’s Car: An AI-powered music generation platform that allows users to create music in the style of famous artists.
  9. Roli Blocks: A music creation system that uses AI to help users create music with a series of blocks.
  10. Soundraw: An AI-powered music generation platform that creates custom music tracks for videos and ads.
  11. Epidemic Sound: A music licensing platform that uses AI to generate music tracks for videos and ads.
  12. MusicVAE: A generative model that uses AI to create music in various styles.
  13. Google’s NSynth: A neural network-powered music generation platform that creates new sounds and music.
  14. WaveNet: A generative model developed by Google that uses AI to generate music and audio.
  15. Brain.fm: An AI-powered music generation platform that creates music for focus, relaxation, and sleep.

These platforms use various AI technologies such as machine learning, deep learning, and neural networks to generate music. They can be used for a variety of applications, including music production, video scoring, advertising, and more.

Here’s a criteria to rate AI-powered music generation platforms:

Criteria 1: Quality of Music Generation

How well does the platform generate music that sounds natural and coherent? How varied and diverse are the generated music tracks? Does the platform produce music that is comparable to human-created music? Rating Scale: 1-5 (1 = Poor, 2 = Fair, 3 = Good, 4 = Very Good, 5 = Excellent)

Criteria 2: User Interface and Experience

How easy is it to use the platform, even for those without musical experience? How intuitive is the interface, and how easy is it to navigate? Does the platform provide a seamless and enjoyable user experience? Rating Scale: 1-5 (1 = Poor, 2 = Fair, 3 = Good, 4 = Very Good, 5 = Excellent)

Criteria 3: Customization and Control

How much control does the user have over the music generation process? Can the user specify parameters such as genre, mood, tempo, and instrumentation? How customizable are the generated music tracks? Rating Scale: 1-5 (1 = Limited, 2 = Some, 3 = Moderate, 4 = High, 5 = Very High)

Criteria 4: Originality and Uniqueness

How original and unique are the generated music tracks? Does the platform produce music that sounds like it was created by a human, or is it easily identifiable as AI-generated? How often does the platform produce novel and unexpected results? Rating Scale: 1-5 (1 = Not Original, 2 = Somewhat Original, 3 = Original, 4 = Very Original, 5 = Extremely Original)

Criteria 5: Output Flexibility

How flexible are the generated music tracks in terms of format and usage? Can the user export the music in various formats, such as WAV, MP3, or MIDI? How easily can the generated music be integrated into other projects or applications? Rating Scale: 1-5 (1 = Limited, 2 = Some, 3 = Moderate, 4 = High, 5 = Very High)

Criteria 6: Collaboration and Integration

How well does the platform integrate with other music creation tools and software? Can the user collaborate with others in real-time using the platform? How seamless is the integration with other creative tools and workflows? Rating Scale: 1-5 (1 = Poor, 2 = Fair, 3 = Good, 4 = Very Good, 5 = Excellent)

Criteria 7: Cost and Accessibility

How affordable is the platform for individual users or businesses? Are there any free or trial options available? How accessible is the platform for users with varying levels of musical experience or expertise? Rating Scale: 1-5 (1 = Expensive, 2 = Somewhat Affordable, 3 = Affordable, 4 = Very Affordable, 5 = Free or Low-Cost)

Criteria 8: Documentation and Support

How comprehensive is the documentation and user guides provided by the platform? How responsive is the support team to user queries and issues? How active is the community and user base around the platform? Rating Scale: 1-5 (1 = Poor, 2 = Fair, 3 = Good, 4 = Very Good, 5 = Excellent)

By using these criteria, you can evaluate and rate AI-powered music generation platforms based on their strengths and weaknesses, and make informed decisions about which platform to use for your specific needs.

Here are the ratings for each of the 16 AI-powered music generation platforms based on the criteria I provided earlier:

Amper Music

Total Score: 33/40

AIVA

Total Score: 37/40

Jukedeck

Total Score: 34/40

Endel

Total Score: 40/40

Magix Music Maker

Total Score: 29/40

Logic Pro X

Total Score: 38/40

Flow Machines

Total Score: 31/40

Daddy’s Car

Criteria 1: Quality of Music Generation - 3 (Good) Criteria 2: User Interface and Experience - 4 (Very Good) Criteria 3: Customization and Control - 3 (Moderate) Criteria 4: Originality and Uniqueness - 2 (Somewhat Original) Criteria 5: Output Flexibility - 3 (Moderate) Criteria 6: Collaboration and Integration - 3 (Good) Criteria 7: Cost and Accessibility - 3 (Affordable) Criteria 8: Documentation and Support - 3 (Good) Total Score: 28/40

Roli Blocks

Criteria 1: Quality of Music Generation - 4 (Very Good) Criteria 2: User Interface and Experience - 5 (Excellent) Criteria 3: Customization and Control - 4 (High) Criteria 4: Originality and Uniqueness - 3 (Original) Criteria 5: Output Flexibility - 4 (High) Criteria 6: Collaboration and Integration - 4 (Very Good) Criteria 7: Cost and Accessibility - 3 (Affordable) Criteria 8: Documentation and Support - 4 (Very Good) Total Score: 34/40

Soundraw

Criteria 1: Quality of Music Generation - 4 (Very Good) Criteria 2: User Interface and Experience - 4 (Very Good) Criteria 3: Customization and Control - 4 (High) Criteria 4: Originality and Uniqueness - 3 (Original) Criteria 5: Output Flexibility - 4 (High) Criteria 6: Collaboration and Integration - 3 (Good) Criteria 7: Cost and Accessibility - 4 (Very Affordable) Criteria 8: Documentation and Support - 4 (Very Good) Total Score: 34/40

Epidemic Sound

Total Score: 33/40

MusicVAE

Total Score: 37/40

Google’s NSynth

Total Score: 39/40

WaveNet

Total Score: 39/40

Brain.fm

Total Score: 33/40

Sono

Total Score: 33/40

Here is the ranking of the platforms based on their total scores:

  1. Endel - 40/40
  2. Google’s NSynth - 39/40
  3. WaveNet - 39/40
  4. Logic Pro X - 38/40
  5. MusicVAE - 37/40
  6. AIVA - 37/40
  7. Jukedeck - 34/40
  8. Roli Blocks - 34/40
  9. Soundraw - 34/40
  10. Amper Music - 33/40
  11. Brain.fm - 33/40
  12. Epidemic Sound - 33/40
  13. Sono - 33/40
  14. Flow Machines - 31/40
  15. Daddy’s Car - 28/40
  16. Magix Music Maker - 29/40

Sono id only the 13 place among 16 models, scoring 33/40.

I am going to try to prompt a rock n roll using Sono in this blog. Here’s the procedure:

flowchart LR
  A[Prompt lyrics] --> B[Separate lyrics into verses and bridges]
  B -->C[Prompt the first verse]
  C -->D[Choose the better version between two generated by the platform]
  D-->E[Prompt the second based on the first and lyrics]
  E-->F[Choose the better version between two generated by the platform]
  F-->G[Prompt the bridge based on the first, second verse and lyrics]
  G-->H[Choose the better version between two generated by the platform]
  H-->I[etc.]
  I-->J[Complete the song]

rnr.jpg

First, we prompt alloma to generate lyrics.

prompt: Generate lyric of a song based on this paragraph: >Drugster, dancer in club, with inpure relationship between several men, Mimi seems as a typical gang girl in 1990s New York. However, when she actually met the love, she was frank and unfaired. Though she was too sexy for her age of 19, she still hold on to the purity of love. When Roger suspected her and Benny, she bluntly asked trust from her lover. It was the truth that her struggled in drugs and disease, but as she said, Its a living.

Christine and Mimi were quite other girls as the similar age in entirely different society. It is fun to be aware of the conventional girl maid in literary and artistic works who always seeks for love and hold her purity regardless of the difficults and hits of life. What’s more, girls are not always as pure as white paper. To Christine, the contact with fantom was secret and finally shocked the surrundings. To Mimi, she was the embodiment of girlhood rebel and crime.

Here is a song lyric based on the provided paragraph:

(Verse 1) In the heart of New York City, in the 90s scene A drugster, a dancer, a girl named Mimi With many men, in impure relations She seemed like a typical gang girl in the nation

But when love found her, she was frank and fair With Roger, she asked for trust, without a care Though young, just nineteen, she held onto love’s purity In the midst of drugs and disease, she lived her reality

(Chorus) Mimi, a living embodiment Of a girlhood rebellion, a crime representation Christine, a literary figure In a different world, seeking love, pure and true

(Verse 2) Christine, a maiden in a different time In literature and art, her love did shine Secret trysts, a phantom in her life Shocking society, igniting strife

Not always pure, not always white Girls like Mimi, living their plight In the heart of New York City, in the 90s scene A drugster, a dancer, a girl named Mimi

Then I tried to put the whole lyric into the prompt but turned out that the platform can only complete the first six lines of the lyric, making a song of roughly 30 seconds. To better prompt, I separated different parts of the lyric.

I described as: rock and roll style of the 80s.

After generating the first verse, two versions are provided, and I chose the better one with the criteria: 80s style and coincides the lyrics.

Then I generated the second and the rest based on the same procedure.

Here’s the final work:

Audio('Heroines%27%20Lament.mp3')