Observations After ~1 Month Using Suno 5/5.5 (Possible Breakthrough?)
I’ve been using Suno pretty heavily for about a month now, and I think I may have stumbled onto something that helped solve a problem I kept running into. My songs often sounded too rigid or “AI-flat,” especially in the vocals.
I want to be clear up front. I’m not an expert and this isn’t anything official. This is just observational trial and error from generating a lot of tracks and trying to understand why some worked and others didn’t.
What I started noticing is that the Style field seems to dominate much more than expected. The more I added to it, like genres, instruments, or mood, the more the song would lock in and ignore subtle instructions in the lyrics. This was especially noticeable with vocal phrasing, dynamics, and spacing.
So I tried something simple. I removed most of the Style field, sometimes all of it, and started encoding everything into the lyrics. Structure, mood, dynamics, even hints of genre.
The result was surprisingly consistent.
The songs became more dynamic. They responded better to vocal shaping. Spacing actually mattered. The phrasing felt more intentional. Overall, they sounded less like a template and more like something performed.
In a lot of cases, they honestly started sounding more human. Not because they were more complex, but because they had space, variation, and better vocal behavior.
It feels like when Style is reduced, the model starts listening more to the lyrics as a control system instead of defaulting to a predefined pattern.
Again, this is just what I’m seeing, but it has been consistent enough that I’ve changed how I prompt entirely.
Here is a look at what I put in my lyrics and style section, keeping Weirdness around 65% and Style Influence 35%. You get a lot of vocal range and ups and down, and you can actually kind of control it by doing this. I found that adding the styles (like this one is a mix of genres) in the description will weight it more, but you lose some of touches you would put in the lyrics, whereas empty or just vocal style helps lift the directions of what you put into lyrics, but you will get some fun randomness.
Example Style example (minimal on purpose):
expressive vocal, wide dynamic range, intimate to powerful, emotional tone
Lyrics format (just an example of how I would generate it)
[Intro | ambient cinematic | distant pads, reverse textures, no drums | extremely sparse | establish space | 24 bars | very low energy | breath and air present]
(whisper, fragile, close mic, almost spoken)
looove..............
i_hear_you_in_the_siiilence..............
dooon’t_move..............
[Verse 1 | minimal piano + sub bass | slow tempo feel | wide spacing between lines | allow notes to decay fully | no percussion | intimate tone | 32 bars]
(soft, controlled, low register)
we stayed where the light runs out...
nothing but shadows and sound...
(very light lift)
you said it was over...
but your voooice stayed_clooose..............
[Transition Groove | introduce pulse | soft kick + bass groove | no vocals | let rhythm establish | repetition encouraged | 16 bars]
[Verse 2 | neo-soul / R&B feel emerges | light drums, syncopation | smooth phrasing | slightly tighter spacing | 32 bars]
(warm, slightly more forward tone)
now the night starts to move again...
heartbeat under the skin...
(whisper to tone shift)
do_you_feeeel_it..............
pulling_you_baaack_iiiiin..............
[Genre Shift 1 | Latin groove | nylon guitar, percussion, rhythmic bounce | brighter tone | more movement | 24 bars]
(airy, playful, rhythmic phrasing)
dance_in_the_shadow_light...
slow_turn_and_hold_me_tight...
(legato stretch)
feeeeel_it_rise..............
don’t_let_it_faaaall..............
[Break | strip instrumentation | remove drums | ambient + bass only | silence between phrases | groove memory remains | 16 bars]
(soft, distant, breathy)
stay..............
with_me..............
[Genre Shift 2 | EDM build | rising synths, tension, filter sweep feel | increasing density gradually | no drop yet | 32 bars]
(building intensity, closer phrasing)
hold_me..............
hooold_me..............
don’t_you_leeeet_me_go..............
(energy rising)
we’re getting clooser..............
[DROP | EDM full energy | heavy bass, sidechain feel | rhythmic vocal chops allowed | high energy contrast | 24 bars]
(punchy, rhythmic delivery)
move_now
don’t_stop
feel_it_hit
right_now
[Genre Shift 3 | Rock / electric guitar driven | strong drums, driving rhythm | vocal power increases | 32 bars]
(full voice, gritty edge, forward tone)
we crash through the niiiight..............
fire in every liiiine..............
(aggressive sustain)
don’t_you_break_thiiis_tiiime..............
[Breakdown | remove drums again | guitar sustains, ambient bleed | space returns | contrast reset | 16 bars]
(soft again, emotional drop)
why_did_we_ruuun..............
why_did_we_faaaall..............
[Final Build | combine elements | gradual layering: pad + guitar + rhythm | rising register | emotional tension | 32 bars]
(ascending, urgent)
i_can_feeeel_it_chaaange..............
i_can_hear_our_naaaames..............
(closer, more intense)
we’re not the saaaame..............
[FINAL PEAK | extended climax | full arrangement | choir backing, layered vocals | maximum intensity | allow sustain to fully resolve | 64 bars]
(powerful, open, high register, vibrato allowed)
loooove_meeeee..............
staaay_with_meee..............
taaaake_me_hiiiiigher..............
dooon’t_stoooop..............
(ultimate sustained line, full release)
nooooooo_oooooooooooooooooooooooooooooooooooooooooooooooooooo..............
[Outro | return to ambient | strip everything back | echo of intro | emotional resolution | 16 bars]
(whisper, fading, distant)
looove..............
just_loooove.............