writes:I recently became aware of an AI tool called Sonauto v2, which is capable of creating music. So I decided to try it out.
Martial #1
I asked it to create lyrics for a song about miniature wargaming, mentioning battles, generals and heroism. It gave me this:
(Verse 1)
Dust in the air, a distant drum call
Scanning the field, before we stand tall
Lines in the sand, a tactical choice
Whispers of plans, a commanding voice(Chorus)
Charge, thunder, and steel!
The trumpets peal!
History's stage, where legends appeal
Each roll of the dice, the victory we steal(Verse 2)
Cannons roar out, a shiver of dread
Chess pieces move, where heroes have bled
Swords crash and clang, a glorious sound
Valiant men fall, on hallowed ground(Chorus)
Charge, thunder, and steel!
The trumpets peal!
History's stage, where legends appeal
Each roll of the dice, the victory we steal(Bridge)
From ages past, our spirits they rise
Under the sun, beneath ancient skies(Chorus)
Charge, thunder, and steel!
The trumpets peal!
History's stage, where legends appeal
Each roll of the dice, the victory we steal
I then told it I wanted martial-type music, then asked it to enhance my prompt. It suggested:
Thunderous cavalry charges, echoing war drums, clashing swords and shields, triumphant brass fanfares, intense battlefield ambiance, strategic planning whispers, distant cannon fire, dynamic tempo shifts, epic orchestral arrangements, immersive stereo spread, gritty and powerful instrumentation, tension-filled moments, historic battle reenactment, detailed sound design, commanding presence, grand and cinematic production quality.
I added "deep male voice."
Then I asked it to generate the song.
(All the songs come out as MP3 files lasting 1.35 minutes.) You can hear it here:
Rather amazing that an AI can do that!
Martial #2
I wanted to try again, but this time, I decided to lightly edit the lyrics.
I changed "scanning the field" to "scanning the tabletop" – more wargamey, right?
I changed "Chess pieces move, where heroes have bled" to "Miniatures move, where heroes have bled".
I changed "Each roll of the dice, the victory we steal" to "Each roll of the dice, the victory we seal"
Leaving all other settings unchanged, I hit the generate button and got version #2:
The first part of the sound songs great, but the AI screws up the lyrics and spews non-words. And there's no apparent way to tell the AI to modify the result. You can only "roll the dice" by hitting generate again.
Martial #3
Last time, the AI got screwed up when it hit the word 'tabletop', so I changed the lyric to 'scanning the table'.
I also modified the chorus from "Each roll of the dice, the victory we seal" to "With each roll of the dice, the victory is sealed".
And the third version:
Epic
Now I decided to experiment with the text prompt for the music, leaving the lyrics unchanged. The AI seems reluctant to copy a specific group or artist, but if you say 'in the style of [blank]' and ask the AI to expand your prompt, it will do so.
So this time, I said 'in the style of the rock band Queen', and the generated prompt was:
Epic stadium anthem, soaring vocal harmonies, powerful guitar riffs, dynamic drum fills, grandiose piano flourishes, theatrical flair, energetic tempo shifts, larger-than-life production, bombastic crescendos, rich and full-bodied sound, confident and charismatic performance, classic rock influences, majestic and triumphant atmosphere.
And the resulting song was:
Intimate
I suggested 'in the style of a 19th Century folk song', and the generated prompt was:
Intimate parlor setting, flickering candlelight, rustic wooden floorboards creaking softly, delicate fingerpicked acoustic guitar, heartfelt vocals with a touch of melancholy, occasional harmonica accents, warm analog recording quality, vintage crackle and warmth, evoking nostalgia and simplicity, capturing the essence of a bygone era with authenticity and charm.
I added 'Irish'.
And the resulting song was:
Note that the AI has trouble with the first iteration of the chorus, but the other iterations are fine.
Majestic
I suggested 'in the style of the great Renaissance composer George Frideric Handel', and the generated prompt was:
Grandiose orchestral arrangement, majestic brass fanfares, intricate string counterpoint, regal harpsichord embellishments, soaring vocal melodies, dynamic tempo changes, Baroque-inspired ornamentation, rich harmonic textures, dramatic crescendos and diminuendos, stately and dignified mood, intricate and ornate composition, studio quality recording capturing the depth and clarity of each instrument.
And the resulting song was:
Soaring
I suggested 'in the style of the Tabernacle Choir at Temple Square' (formerly known as the Mormon Tabernacle Choir), and the generated prompt was:
A grand cathedral performance, soaring vocal harmonies, powerful choir dynamics, majestic pipe organ accompaniment, resonant brass section, angelic soprano solos, reverberant acoustics of a vast sacred space, rich and full-bodied timbre, emotive crescendos, divine atmosphere, meticulously mixed for clarity and presence.
And the resulting song was:
Mystical
I suggested 'in the style of Enya', and the generated prompt was:
A mystical journey through ancient ruins, ethereal choir-like vocals, haunting Celtic harp melodies, soft orchestral swells, distant echoes of forgotten chants, shimmering synth pads, gentle percussive heartbeat, slow and mesmerizing tempo, expansive soundstage with a cathedral-like reverb, dreamy and otherworldly ambiance, rich and lush production quality, evoking a sense of wonder and mystery.
And the resulting song was:
Surf
I suggested 'in the style of The Beach Boys', and the generated prompt was:
Vibrant surf scene at sunset, crashing waves against golden sands, shimmering electric guitar riffs, harmonious vocal harmonies, energetic drum beats, lively bassline, upbeat tempo, sunny and carefree vibe, vintage pop influence, polished production with a retro touch, beach party ambiance, bright and colorful sound palette, wide stereo spread, reverb simulating seaside acoustics, nostalgic and joyful mood.
And the resulting song was:
Reverent
I suggested 'in the style of a Gregorian Chant', and the generated prompt was:
Ancient cathedral echoes, reverent harmonies rising, deep male voices intoning, ethereal female soprano soaring, resonant chamber acoustics, hauntingly beautiful melodies, sacred and meditative aura, slow and steady tempo, immersive and spiritual ambiance, rich texture of layered voices, transcendent and timeless quality, produced for clarity and depth, divine and otherworldly sound experience.
And the resulting song was:
Poprock
I suggested 'in the style of The Beatles', and the generated prompt was:
Crafting a keyword prompt for audio generation based on The Beatles: Innovative 60s pop-rock vibes, harmonious vocal melodies, jangly electric guitars, groovy basslines, lively drum beats, catchy hooks, nostalgic retro sound, dynamic song structures, warm analog textures, energetic and uplifting mood, studio quality production, vibrant stereo spread, iconic British Invasion influence, timeless musical craftsmanship.
And the resulting song was:
Coffeehouse
I suggested 'in the style of James Taylor', and the generated prompt was:
Intimate coffeehouse performance, soulful acoustic guitar strums, warm and inviting vocals, gentle folk influences, nostalgic storytelling lyrics, cozy and familiar vibe, laid-back tempo, subtle fingerpicked melodies, rich and organic sound, studio quality recording, authentic and heartfelt delivery, evoking a sense of comfort and introspection.
And the resulting song was:
Baroque
I suggested 'in the style of Johann Sebastian Bach', and the generated prompt was:
Intricate baroque cathedral, grand pipe organ resonating, intricate harpsichord arpeggios, delicate violin counterpoint, rich cello bass lines, precise contrapuntal melodies, ornate ornamentation, regal and majestic atmosphere, classical elegance, dynamic phrasing, crisp articulation, authentic period instruments, studio quality recording, immersive stereo spread, reverberant acoustics reminiscent of a historic concert hall.
And the resulting song was:
Funky
I suggested 'in the style of a 1980s TV theme song', and the generated prompt was:
Funky wah-wah guitar riffs, groovy bassline, lively brass section, energetic drum fills, soulful Hammond organ melodies, upbeat tempo, dynamic horn arrangements, nostalgic retro vibe, vibrant and colorful instrumentation, tight rhythm section, catchy hooks, high-energy feel, polished production with vintage warmth and character.
And the resulting song was:
Soulful
I suggested 'in the style of Ray Charles', and the generated prompt was:
Soulful piano chords, expressive brass section, groovy rhythm section, emotive gospel-inspired vocals, dynamic call-and-response exchanges, rich Hammond organ fills, energetic uptempo swing, lively horn arrangements, vibrant and engaging performance, warm vintage analog sound, authentic and heartfelt delivery, polished production with a touch of rawness, capturing the essence of classic rhythm and blues.
And the resulting song was:
You may say that AI-generated music is soulless, but I am amazed that AI can do this at all. What do you think?