This is a hypothetical business case by SasaDaily.
A handmade food brand with only four team members
regularly produces:
new product Reels,
process videos,
seasonal product clips,
and customer unboxing content each week.
The most troublesome part isn’t always:
shooting,
nor editing,
but at the end of video production,
someone suddenly asks:
“Where’s the music?”
Then one person starts:
browsing music libraries,
trying one track,
switching to another,
trimming it to 45 seconds,
realizing the climax comes too early,
then changing again.
A video that could have been done in 30 minutes
ends up taking an extra:
20 to 30 minutes
just searching for
a "good enough" background track.
This is exactly what the company wants to improve.
They didn’t start by asking, “Can AI compose songs for us?”
Instead, the team asked:
“Which tasks are repetitive?”
The answer was clear:
every video requires
searching music,
testing tracks,
trimming,
and checking if the rhythm fits.
Thus, Lyria 3.5 serves in this company as
not a composer,
nor a brand director,
but more like a
Music Draft Generator.
A tool for generating first versions of music drafts.
Step 1: The brand decides upfront “the sound we want”
The team doesn’t simply open Gemini and say:
“Make me a nice song.”
Instead, they create a
Brand Music Card
laying out a few core principles, for example:
Overall vibe:
warm,
natural,
handmade feel.
Common instruments:
acoustic guitar,
soft piano,
light percussion.
For typical product videos:
prioritize instrumental music,
avoid overly dramatic styles,
avoid intense EDM,
don’t mimic any particular artist or band.
This step is crucial.
If AI chooses the brand sound
video by video,
today it might be jazz,
tomorrow heavy rock,
the day after synthwave—
each track individually fine,
but collectively inconsistent with the brand identity.
AI can create many directions, but brand consistency remains human-led
Google DeepMind’s Lyria 3.5
allows control over:
genre,
tempo,
instruments,
dynamics,
vocal style,
and can even generate fitting music based on images.
The more capabilities,
the greater the need to narrow scope first.
This company asks not,
“What can Lyria do?”
but rather,
“Within what boundaries do we allow it to vary?”
This is one of the biggest differences between
using AI purposefully for branding,
and just randomly playing with AI.
Step 2: Divide each video’s timeline into three segments
For example, for a 60-second new product video,
where footage is ready:
0–10 seconds:
first product appearance,
10–50 seconds:
production, details, cutting, packaging,
50–60 seconds:
finished product and purchase info screen.
The team doesn’t need
to know composition,
just define the
energy flow for the three sections:
Opening: simple.
Main section: gradually increasing.
Ending: clean fade out.
This is the “three-segment energy flow” taught in today’s one-minute lesson.
Step 3: Use product photos as part of the music brief
Lyria 3.5 can accept images
as contextual input.
So the team can provide
representative product photos
without needing to describe from scratch
“What vibe does this product have?”
They add short text instructions such as:
warm, handmade, natural, instrumental, around 60 seconds, acoustic guitar, soft piano, simple opening, increasing rhythm in the main section, natural ending.
The photos provide:
mood,
the text gives:
direction,
and the timeline offers:
movement.
This is much clearer than simply saying “Make music for my product”
Because “food videos” can have
many musical styles, such as
fine dining,
cute desserts,
street food,
home cooking,
health food,
or festive gift packs,
all of which sound very different.
AI doesn’t know your brand’s true identity,
so it’s best not to expect AI to guess, but
to tell AI your brand’s choices upfront.
Step 4: Create 2–3 drafts per video; no need for first draft to be final
Lyria produces varied outputs;
even with the same prompt,
each generation can differ.
This company’s process is not:
generate → use immediately,
but instead:
generate A, generate B, generate C,
then choose the version
that best fits the video.
When selecting music, they don’t ask “Which sounds best?” first
Because what sounds best
isn’t always the
best fit for the video.
For example: A sounds beautiful alone,
but its opening is too full.
B has simpler melody,
so it doesn’t overpower product sounds, narration, or captions.
The company chooses B,
viewing music as
supporting the content,
not competing with it.
Step 5: When the video includes narration, the music plays a subtler role
If the brand owner says in the video,
“This jam uses three kinds of seasonal fruit,”
the music shouldn’t be complex.
The team adds to the prompt:
instrumental,
supportive under spoken narration,
avoid overpowering voiceover.
If the video has no narration,
just footage of production, food close-ups, cutting, packaging,
music can carry more emotion.
Same tool, different roles.
Step 6: Lyria drafts the music; editing software does precise timing
The company doesn’t demand Lyria to insert a drum at exactly 7.3 seconds,
switch sections perfectly at 31.8 seconds, or finish at 58 seconds.
Music generation isn’t the same as DAW automation.
They only ask for the big energy direction to be correct:
low energy at the start,
increasing in the middle,
fading at the end.
Precise syncing to beats happens later in editing software,
by trimming, moving, and fading tracks accordingly.
This approach is more practical.
AI creates the raw materials; humans make the final editing decisions
This workflow’s key boundary is that Lyria is not a
one-click full content production tool.
Instead, it frees the team from
having to search a huge music library
every time.
Before:
video → find ready-made music → try multiple tracks → trim length → reluctantly fit.
Now:
video → define energy flow → Lyria generates drafts → select best version → editing software fine-tunes timing.
The work order has changed.
Step 7: Official brand jingles require different processes than standard Reels
When the owner likes an AI-generated draft and suggests,
“Let’s use this as our brand’s main theme,”
the team says no,
because
background music for social media
and
official brand assets
have very different risk levels.
Official brand music might be used for many years,
ads, events, across platforms,
and might require exclusivity,
which requires a separate rights review.
SynthID can identify AI-generated audio but can’t handle rights decisions
Google adds SynthID to Lyria’s audio,
which helps identify
whether content was generated or edited by Google AI.
But as discussed today, SynthID mainly manages
source verification,
not
copyright certification
or unlimited commercial use licenses.
The team divides music into two types.
Category 1: Daily content drafts
Such as:
ordinary product Reels,
internal proposals,
unreleased prototypes.
The main goal here is:
speed,
correct direction,
and reducing time spent finding music.
Category 2: Official brand assets
Such as:
brand jingles,
major ads,
long-term opening music,
commercial releases,
large campaigns.
These require extra steps:
rights review,
preserving prompts, versions, manual edits, original raw materials, lyrics, recorded vocals, arrangement changes, and project files.
Professional legal and music rights confirmation is done if necessary.
The company doesn’t avoid AI music due to copyright fears
That would be an extreme position.
Mature approach is to
match review depth to usage intensity.
A social draft used for a few days
and a theme song used for 10 years
don’t require the same process.
AI speeds up music creation,
but companies must implement usage tiering.
Step 8: Avoid prompts that copy specific artists’ styles
The team’s Brand Music Card
explicitly states:
no specific artist names in prompts.
Google’s Lyria Safety Filter
already blocks requests for
particular artist voices or copyrighted lyrics.
A better approach is describing
slow, warm, acoustic, soft female vocal, clean guitar, light percussion,
focusing on music characteristics rather than imitating someone.
Step 9: Each month, track and keep the three best-performing music directions
The company avoids having Lyria guess from scratch every day.
After a month, they find three commonly used styles:
A: acoustic guitar + piano, suited for new products,
B: light funk + bass, suited for process videos,
C: soft ambient, suited for brand stories.
Next month onward, they start generation from
A, B, and C,
not from the entire world of genres.
The AI workflow becomes faster with use.
This is how AI becomes a workflow, not just a chat subject every day
If you re-think prompts daily, try new genres, and repeatedly debate brand feeling,
despite Lyria’s speed, the team wastes time.
A mature company accumulates
Brand Music Cards,
favorite prompts,
energy templates per video type,
and selected versions,
which gradually become the company’s
AI Creative SOP.
Some hypothetical numbers
"SasaDaily’s assumptions."
Assuming this company produces
5 short videos weekly,
4 weeks per month,
totaling 20 videos monthly.
Previously, each video took about
25 minutes
to search, test, and trim music.
20 × 25 minutes =
500 minutes,
roughly 8.3 hours a month.
After switching to Lyria 3.5
"SasaDaily’s assumptions."
Assuming they have Brand Music Card, three common styles, and energy templates established,
each video takes about
8 minutes
to generate drafts, compare, and select.
20 × 8 minutes =
160 minutes,
about 2.7 hours per month.
The theoretical monthly saving is about
5.7 hours.
Internal time costs
"SasaDaily’s assumptions."
Assuming content managers cost NT$650/hour,
5.7 hours × 650 =
approximately
NT$3,700.
But this isn’t
"Lyria generates NT$3,700 extra per month."
Instead,
about 5.7 hours can be reused for
filming additional videos,
building product pages,
responding to clients,
improving packaging,
and analyzing content performance.
This is what we call
time value.
This company doesn’t chase “how many songs AI can generate monthly”
That measures tool usage,
not business results.
What to track are:
1. Average time spent searching music per video. No drop means no time saved.
2. Percentage of generated music actually used. If only one in ten is chosen, the prompt or Brand Music Card may be flawed.
3. Number of music edits per video. Constant re-edits indicate unclear energy direction.
4. Brand consistency. Watching 10 videos in a row should not feel like 10 different companies.
5. Whether official brand assets undergo rights review to avoid quick drafts becoming high-risk assets.
No need to immediately measure traffic changes
This workflow’s first goal isn’t
increasing views.
It solves
production costs.
If previously, only two videos could be done each day,
and now three can be completed in the same time,
that’s operational improvement.
Which music actually improves view duration, completion, or conversions
is a later test, not to be confused with cost reduction.
AI music can also serve as an A/B creative variable
For the same product video with identical footage,
change only the music mood, such as
A: warm acoustic,
B: more rhythmic light funk,
and measure completion rate, dwell time, shares.
AI music can not only save production time,
but also make expensive creative tests more feasible.
Note: only change one variable at a time when testing,
otherwise changing footage, text, music, and CTA all at once
makes it impossible to attribute what caused differences.
Same benefits apply to solo creators
Even if you don’t have a 4-person team,
and you shoot, edit, and post alone,
Lyria is even more practical.
For a solo creator,
the biggest cost is often
their own time.
Saving 15 minutes of music search per video may seem small,
but 20 videos per month equals
5 hours saved.
This is an easily overlooked value of AI tools,
not as a one-shot mega-project killer,
but as reducing daily repetitive tasks.
But don’t let faster generation create unnecessary content overload
AI’s biggest pitfall is that,
when production costs drop,
people just keep creating.
Previously, 5 videos a week was standard,
now thinking AI is fast, they force 25 videos,
resulting in no strategy, no audience, and more team exhaustion.
This company still plans weekly content needs first,
using Lyria only to
reduce music production costs,
not to decide how much content to produce.
True ROI isn’t “AI is faster than people”
It’s whether tasks that weren’t worth significant time
can now be done at lower cost.
Small brands often don’t have
music supervisors,
full-time composers,
or audio editors,
yet need to produce ongoing social videos.
Lyria 3.5 fills the gap
between professional music production and just picking a free track.
It first creates drafts that fit the task,
letting humans decide what to use.
Does this mean professional musicians lose business?
Not really.
If the brand later wants
large campaigns,
sonic branding,
official jingles,
real human vocals,
and long-term music identity,
the company will be clearer on their needs,
having already tested many directions quickly with AI.
When handing over to pros, they can say:
“We found that this rhythm, these instruments, and this energy curve best fit our brand.”
AI lowers exploration costs,
allowing professional work to focus
on truly valuable areas.
This is the complete process for this 4-person company
Weekly content plan
↓
Shoot video
↓
Edit rough length
↓
Pick a representative photo
↓
Apply Brand Music Card
↓
Write
opening/main/ending
energy flow
↓
Lyria 3.5 generates 2–3 drafts
↓
Humans select the best fit for brand and video
↓
Edit software aligns precisely to beats
↓
Watch full playback
↓
Standard social content:
publish according to current rules
↓
Official ads/jingles/long-term brand music:
conduct separate rights review
↓
archive prompts, versions, manual edits
AI does not take over the brand
AI only takes on:
“making several music versions to try out first.”
The brand still decides:
who they are,
how they want to sound,
the video’s emotional needs,
which track fits best,
what can be publicly released,
and which uses require full rights review.
This kind of division of labor
is easier for long-term use
than “AI does everything for me.”
The greatest value of Lyria 3.5 for small brands is not “replacing musicians”
But
for many everyday short videos
that aren’t worth hiring a professional composer for each time,
the team used to have only two options:
music libraries,
or randomly picking tracks.
Now they have a middle option:
AI-generated music drafts that fit each project.
High frequency,
low risk,
fast content.
More valuable, long-term, brand asset music
still incorporates
human creativity,
professional production,
and rights clearance.
This is what a mature AI workflow looks like
Not:
handing AI all tasks,
but:
breaking tasks down.
This company breaks work into:
brand direction:
human,
video needs:
human,
first music draft:
AI,
music selection:
human,
precise editing:
human + software,
rights decisions:
human,
final release approval:
human.
AI only steps in
where it suits best.
What this case really saves
Not:
“We no longer need music.”
But:
previously having to re-search through thousands of ready-made tracks
to find an acceptable one.
Now you say:
how long the video is,
what emotion it needs,
your brand’s sound,
opening style,
middle progression,
ending style,
and AI generates
several comparable versions.
Humans save time for
making real judgments.
This is
AI giving time back to people.
Today, grow a little with AI.
Learn one AI tip every day.
Save a little time each day.
Improve one skill step by step.
SasaDaily grows with you.
Recommended reading
AI can mimic a style in seconds, but creators may have taken ten years to get there
