Written short, not cut short
A :60 truncated to :30 is not a :30. A shorter cut is written to its own word budget: a :30 holds about 51 sung words, and every digit of a phone number is one of them.
Songs, jingles and station imaging
Music that sings your words. At the length you asked for.
Describe the track you want and get finished audio back: a song or an instrumental bed, an advertiser's jingle with the phone number sung in it, or a station's imaging cut to the slot. Anything from :10 upwards, written to the length rather than trimmed to it, and you can hand it your own lyric instead of a description.
- 10 min
- the engine's longest render, and :10 its shortest. Your plan sets the ceiling inside that
- 21
- jingle and imaging archetypes, or the blank page a song starts from
- $99
- a month for 13 hours of audio, stems and tracks up to 5:00
What it does
Finished audio, with the numbers to prove it.
We run the music engine ourselves, so nobody takes a cut of every render and what you make is priced by the second rather than by the song. Every card below names a real number rather than an adjective.
Describe it, or hand it your words
Three ways in, and which one you get is decided by what you send. A description, and the words are written before anything renders. Your own lyric, and those exact words are sung. An instrumental flag, and nothing is sung at all. Only the first needs a language model configured.
One bed, 6 slots
A package is one musical bed timed to the 6 slots that cover a broadcast day: :10 between records, :15 into the news, :30 under the morning show. 6 renders that sound like one station.
Billed by the second
The unit is a second of finished audio, never a song. A sung :15 sweeper costs 19 seconds of the allowance and a 3 minute song costs 225, on one rate card. Nothing here is priced per track, so a short piece is cheap rather than a whole credit.
The number is a deliverable
The must-say list is checked against the finished words before anything renders. A draft that dropped the phone number goes back with the omission named, and it goes back to the writer rather than to you.
Section markers survive
Write [verse] and [chorus] in your lyric and the arrangement follows them. The engine sings any marker it does not know, so what the interface calls a Tag goes on the wire as [chorus] rather than putting the word in the vocal.
The frequency, every time
The dial position is matched as digits, so a line sung one oh one point five satisfies a profile that says 101.5. Twenty sweepers read from one profile, so there is no twentieth chance to get it wrong.
Make that one again
Every take carries the seed it came from, the caption that was actually sent, and the model that rendered it. The take you kept comes back next month sounding like itself, whether the person who approved it was a client or you.
A style you can save
A persona is a recipe rather than a likeness: the tags, the tempo, the key and the production notes that got you the sound you kept. Station holds 40 of them and Station holds 40.
Some briefs do not fit
A toll free number is 11 sung words. The form says a brief will not fit at :15 before a GPU second is spent, and offers the length that holds it.
Pull the bed out from under it
Stem export separates vocals, drums, bass and the rest, so a voice can read over the same bed the sung version is built on, or a video can take the instrumental. One pass, however many files come back.
Longer, without paying twice
Extending a finished track is billed on the seconds it adds and never on the whole result. The one thing you already paid for is the part you keep.
One lyric, three arrangements
Cover a track you own in another style and the words and the shape stay put: one set of lyrics as country, as pop and as rock is three finished tracks rather than three writing jobs. For a rep it is also three stations off one client approval.
A draft is not a render
Writing words runs a language model rather than a GPU, so it is a flat 5 seconds of the allowance however long the lyric and no audio is made. Drafts wait in a library until you send one to the studio, which copies the words onto that track: delete the draft next month and the finished piece still holds what it was sung from.
Cleared to use, in writing
Commercial use on every plan, the free one included, advertising you bill a client for included. There is no tier where what you rendered is yours to hear but not to air, so nothing has to be made twice to be cleared.
No surprise invoice
The allowance is a ceiling, not a meter. When it is spent, renders stop and we ask you to move up a plan. There is no overage line, and no script of yours can run up a bill overnight.
A cold start is not a hang
The engine sleeps when nobody is using it, and the first render after a quiet spell spends 30 to 90 seconds waking up. Every screen says Queued while that happens and Rendering once it does, because otherwise the two look identical.
A room, not a login
Seats and simultaneous renders both come with the plan: Station carries 10 seats and 4 renders at once, Station carries 10 and 4. Concurrency is a card count rather than a throttle, so it is the number that decides how a Friday afternoon goes.
4 things it deliberately will not do.
- Clone a singer. A persona here is a saved style recipe, not a likeness, and we will not call it a voice.
- Master your audio. What comes back is a finished mix at the length you asked for, and your processing chain is still yours.
- Clear music you did not make here. The licence covers what this engine rendered for you and nothing else.
- Render shorter than :10. The engine's floor is 10 seconds, so a :05 stinger is not something this makes today.
How it works
Three steps, whatever you are making.
- 1
Say what it is, or paste the words
A sentence is enough: what it is for, how it should feel, who is singing. Or paste your own lyric and those exact words are sung. An advertiser's brief and a station profile are the same step with fields instead of a sentence, because those two get checked against a list afterwards.
- 2
Pick the length
Anywhere from :10 to 10 minutes, with your plan's ceiling inside that, or one of the 4 broadcast lengths when it has to fit a log. The words are written to the length, and the cost is stated before you press.
- 3
Audition takes, keep one
Ask for four and get four alternates of one piece, under one title, rather than four entries burying the rest of your library. Download the full length audio, or the stems.
Songs and beds
Your words, sung, or no words at all.
One lyric, exactly as the engine receives it, with the syllable cost of every line beside it. Write a description and the words are written for you before anything renders, or paste your own and those exact words are sung. Section markers survive into the arrangement, and an instrumental is a flag on the render rather than a hopeful line in a prompt.
The words, as the engine receives them
Cross Street, opening theme
Opening theme for a two person podcast about cycling across a city at night. Warm indie folk, female lead, acoustic guitar and hand claps, ends on the show name.
- Verse[verse]
- 9Wheels on the wet road, quarter past five
- 12Everybody's heading home, we're heading out
- 10Two helmets, one map and a bad idea
- Chorus[chorus]
- 4Cross Street, Cross Street
- 9Two of us and a city to cross
- 8Cross Street, every Thursday night
- Chorus[chorus]
- 2Cross Street
Budget at :30
69
syllables the length holds
Written
54
syllables in the body above
Sings in about
23.6s
-21% off :30
The shape at each length
Each section shows the syllables it may spend. The figure after the bar is the whole row.
- :1534Chorus34
- :3069Verse32Chorus37
- 1:00138Verse32Chorus37Verse32Chorus37
- 3:00413Verse66Chorus76Verse66Chorus76Bridge53Chorus76
Station imaging
6 renders. One station.
Type the station name, the dial position and the slogan once, and every cut in the package reads from that one record. The bars below are the same bed timed to the slot each piece has to fill, which is what a listener hears when your sweepers agree with each other.
The same bed, cropped to each slot
- :10Shotgun IDFully sung
- :15Station anthemFully sung
- :10Power sweeperFully sung
- :15News themeInstrumental bed, nothing sung
- :10Traffic themeSung tag over a bed
- :30Morning show themeFully sung
Advertiser jingles
The spec spot is finished before the meeting starts.
Business name, offer, phone number, call to action, and any line that has to be said word for word. The lyric is written against that list and checked against it, and only then sung. What you played in the meeting is the file you traffic.
The brief
- businessName
- Valley Tire and Auto
- phone
- 555 0142
- offer
- Two tires fitted free with any four
- cta
- Call today
- mustSayPhrases
- All work guaranteed twelve months
- targetLengthSec
- 30
Checked against the finished words
- RequiredBusiness name: Valley Tire and AutoValley Tire and Auto, out on the Ridgeway road
- RequiredPhone number: 555 0142Call five five five, oh one four two
- ParaphrasedOffer: Two tires fitted free with any fourReshaped in the lyric rather than sung word for word, and reported to you rather than enforced.
- SungCall to action: Call todaySo call today, we're open until six
- SungCity: RidgewayValley Tire and Auto, out on the Ridgeway road
- RequiredMust-say line: All work guaranteed twelve monthsAll work guaranteed twelve months
Sung words each length holds
Filled: what the required facts cost. Tick: the 60% a lyric may spend on facts.
- :1017 wordsRefused at the formneeds a :30
- :1525 wordsRefused at the formneeds a :30
- :3051 wordsFits
- :60101 wordsFits
The same brief, measured against every broadcast length. A length whose bar crosses the tick is refused on the form before a GPU second is spent, and the refusal names the length that would hold the brief instead. A spot that runs :32 cannot go in a :30 slot at any price, which is why a shorter cut is written shorter rather than edited shorter.
Rights and files
What comes back, and what you may do with it.
The same three answers whichever of the three you came for: a finished mix at the length you asked for, the parts underneath it if you want them, and a licence that says in writing you can sell what you made.
Commercial rights
The licence covers what this engine rendered for you: air it, sell it, bill a client for it. It covers nothing else, so a track you brought in to cover or extend is still governed by whatever you hold on that.
Every plan, Free included: commercial
Stems and files
The finished mix at the length you asked for, ending on a button rather than trailing off, so it can butt against whatever comes next: a traffic log cannot take a fade and neither can a hard cut to picture. Stem export on every paid plan: vocals, drums, bass and the rest, so the bed can go under a voice, a narration or a scene.
Your workspace
What you make is stored under your workspace rather than in a shared pool, and nothing you render is offered to anybody else. Your track does not turn up in somebody's feed and the station across town does not get your package, because there is no browse page for either of them to appear on.
Pricing
Priced by the second of finished audio, never by the song.
A sung :30 advertiser spot costs 38 seconds of the allowance and a 3 minute song costs 225, so the Station plan is about 208 songs a month or 1,231 spots, and the Station plan is about 2,463 sweepers at :15. Every figure in this paragraph is computed from the same file the quota gate reads.
Free
Free
Ten minutes of audio a month, to find out whether it can sing your name.
- 10 minutes of rendered audio a month
- Tracks up to 1 minute
- One render at a time
- One saved style persona
- Full commercial and broadcast rights
$19/month
Creator
Two hours a month for one person, stems included.
- 2 hours of rendered audio a month
- Tracks up to 4 minutes
- Two renders at a time
- Stem export: vocals, drums, bass and the rest
- Full commercial and broadcast rights
- 1 seat
$49/month
Studio
For a shop rather than a person: more of everything, and three seats.
- 6 hours of rendered audio a month
- Tracks up to 5 minutes
- Three renders at a time
- Stem export: vocals, drums, bass and the rest
- Full commercial and broadcast rights
- 3 seats
Start here
$99/month
Station
A signal's worth of imaging, and an API for the automation to call.
- 13 hours of rendered audio a month
- Tracks up to 5 minutes
- Four renders at a time
- Stem export: vocals, drums, bass and the rest
- Full commercial and broadcast rights
- Public API and keys
- 10 seats
$199/month
Cluster
Several signals in one market, on one account and one bill.
- 28 hours of rendered audio a month
- Tracks up to 5 minutes
- Six renders at a time
- A station profile per signal, so a package never mixes call letters
- Stem export: vocals, drums, bass and the rest
- Full commercial and broadcast rights
- Public API and keys
- 20 seats
$299/month
Group
Several markets, or an agency running call letters that are not its own.
- 45 hours of rendered audio a month
- Tracks up to 10 minutes
- Eight renders at a time
- Stem export: vocals, drums, bass and the rest
- Full commercial and broadcast rights
- Public API and keys
- 50 seats
Custom
Annual or on your own terms
Enterprise
For a group, a network or a rep firm: the terms are written, not picked.
- 150 hours of rendered audio a month, or whatever the deal says
- Tracks up to 15 minutes
- Sixteen renders at a time
- 250 seats
- Invoiced on your terms rather than on a card
- A written agreement, not this page's standard one
A sung second costs 1.25 seconds of your allowance and an instrumental second costs 1. Stem export costs 0.75 per second of the track it separates, and writing lyrics costs a flat 5 seconds. Every render is billed at least 5 seconds.
What the allowance does when it runs out, and what happens to a render that failed on our side, are stated in full on the pricing page. The short version is on the card above: renders stop, nothing overages.
We do not publish a render time. It depends on the length and on whether the lyric writer ran, and we have not measured it at every length, so it is not on this page until we have. What we will say is that the engine sleeps when nobody is using it, and the first render after a quiet spell spends 30 to 90 seconds waking up before the render starts.
Prices in USD. Commercial and broadcast rights come with every plan, the free one included, and the Terms of Service are what govern.
For developers
Post a brief, poll a song id.
Station and Cluster and Group and Enterprise include the public API and key minting in the console. Post a prompt, a lyric, an advertiser's brief or a station profile, get a song id back, and poll it or take a webhook.
- Keys are issued per workspace and spend that workspace's allowance, whoever made the call.
- Anything that spends takes an idempotency key, because a render can legitimately outlast a client's timeout and a retry must not bill twice.
- The contract at /api/v1/openapi.json is keyless, so read the whole surface before you sign up for anything.
POST /api/v1/songs
curl -X POST 'https://jinglebuddy.radioworkflow.com/api/v1/songs' \
-H "Authorization: Bearer $JINGLEBUDDY_API_KEY" \
-H 'Content-Type: application/json' \
-H 'Idempotency-Key: 6f9619ff-8b86-d011-b42d-00c04fc964ff' \
-d '{
"prompt": "Warm family pizzeria jingle, singable hook, upbeat and friendly",
"kind": "song",
"style_tags": [
"uptempo pop",
"bright horns",
"female vocal"
],
"instrumental": false,
"duration_sec": 30,
"takes": 1,
"is_public": false
}' \
--max-time 900
Hear your own words come back sung.
10 minutes of audio a month with no card, which is about 8 tracks at the free plan's 60 second ceiling, or 46 sung :10 station IDs. Enough to decide, and cleared to air: the licence is the same on the free plan as on any other.