AUG 12, 2026
Flow Music: Generate Songs with Google’s Lyria 3.5
Table of contents
- Generate an instrumental on Lyria 3.5
- Sing your own lyrics
- Editing existing songs
- Which model should you use
Google shipped Lyria 3.5 to Flow Music, describing it as its top-performing model with more dynamic vocals and richer musicality. The Flow Music API v1 now exposes it through a new optional model parameter on POST /music and POST /music/edit.
model accepts lyria-3.5 or lyria-3-pro. lyria-3-pro remains the default. Every song generated through this API since launch was rendered on it, so leaving model out keeps your output exactly as it was — nothing changes for existing integrations unless you opt in.
Generate an instrumental on Lyria 3.5
Add "model": "lyria-3.5" to any POST /music call. Everything else works the same, and each generation still returns up to two clips as an A/B pair.
Analog Tides — warm analog synthwave, instrumental
curl — POST /music (instrumental, Lyria 3.5)
curl --location 'https://api.useapi.net/v1/flowmusic/music' \
--header 'Authorization: Bearer user:1234-…' \
--form 'prompt="warm analog synthwave, steady 100bpm, shimmering arpeggios"' \
--form 'instrumental="true"' \
--form 'model="lyria-3.5"'
Sing your own lyrics
Pass lyrics and the model sings your words instead of writing its own. Tag the sections with [Verse] and [Chorus] to shape the arrangement.
Dashboard Drift — the same synthwave prompt, with supplied lyrics
curl — POST /music (custom lyrics, Lyria 3.5)
curl --location 'https://api.useapi.net/v1/flowmusic/music' \
--header 'Authorization: Bearer user:1234-…' \
--form 'prompt="warm analog synthwave, steady 100bpm, shimmering arpeggios"' \
--form 'model="lyria-3.5"' \
--form 'lyrics="[Verse]
Neon lines across the dashboard glow
[Chorus]
Drive me slow, drive me slow"'
Dashboard Ghost — a slower ballad prompt over a longer lyric
curl — POST /music (ballad, Lyria 3.5)
curl --location 'https://api.useapi.net/v1/flowmusic/music' \
--header 'Authorization: Bearer user:1234-…' \
--form 'prompt="synthwave ballad, warm analog pads, emotive female vocal, 90 bpm"' \
--form 'model="lyria-3.5"' \
--form 'lyrics="[Verse]
Neon rain on the boulevard
I keep your ghost in the dashboard light
[Chorus]
Drive me slow through the afterglow
We were never meant to say goodnight"'
The model expands what you give it — the returned clip.lyrics holds the full lyric it actually sang, which is usually longer than the seed you passed.
Editing existing songs
POST /music/edit takes model too, alongside cover, extend, replace and remix. The four examples below are all edits of one existing track, each requested with model: lyria-3.5.
cover restyles the whole song, here to solo piano.
curl — POST /music/edit (cover)
curl --location 'https://api.useapi.net/v1/flowmusic/music/edit' \
--header 'Authorization: Bearer user:1234-…' \
--form 'operation="cover"' \
--form 'clip="user:[email protected]:…"' \
--form 'instruction="solo piano"' \
--form 'model="lyria-3.5"'
extend adds new music to the end — 30 seconds here, taking the track from 175 s to 205 s.
curl — POST /music/edit (extend)
curl --location 'https://api.useapi.net/v1/flowmusic/music/edit' \
--header 'Authorization: Bearer user:1234-…' \
--form 'operation="extend"' \
--form 'clip="user:[email protected]:…"' \
--form 'instruction="build to a big finish"' \
--form 'extend_s="30"' \
--form 'model="lyria-3.5"'
replace regenerates one section in place, leaving the rest untouched. This one swaps seconds 20 to 40 for a bright synth lead.
curl — POST /music/edit (replace)
curl --location 'https://api.useapi.net/v1/flowmusic/music/edit' \
--header 'Authorization: Bearer user:1234-…' \
--form 'operation="replace"' \
--form 'clip="user:[email protected]:…"' \
--form 'instruction="swap in a bright synth lead"' \
--form 'mask_start_s="20"' \
--form 'mask_end_s="40"' \
--form 'model="lyria-3.5"'
remix re-sings the song with edited lyrics over the original sound. Pin seed to make the run reproducible.
curl — POST /music/edit (remix)
curl --location 'https://api.useapi.net/v1/flowmusic/music/edit' \
--header 'Authorization: Bearer user:1234-…' \
--form 'operation="remix"' \
--form 'clip="user:[email protected]:…"' \
--form 'prompt="warm analog synthwave, steady 100bpm, shimmering arpeggios"' \
--form 'seed="777"' \
--form 'model="lyria-3.5"' \
--form 'lyrics="[Verse]
Neon lines across the dashboard glow
[Chorus]
Drive me slow, drive me slow"'
effect and stems accept model as well, but their output is unaffected — those are signal processing and stem separation, with no music model in the loop.
Which model should you use
Both models are available on every Flow Music plan, including free accounts, and cost the same per song. lyria-3-pro stays the default, so nothing changes until you ask for something else. If you want the newer model, send "model": "lyria-3.5" and compare against your own prompts — that is the only comparison that matters for your material.
See POST /music and POST /music/edit for the full parameter reference, or the Flow Music API overview for pricing and the cost calculator.