AI and Soca: What to Check Before Carnival 2027

The short version

Timothy Temal forecast 60–80% AI use in 2027 Carnival productions during an interview. An independent industry count is still missing from this review.
Cleaning up a recording and generating new music raise different creative and permission questions. Name the task before choosing a tool.
Trinidad and Tobago's Intellectual Property Office says copyright protection does not depend on registration. COTT's work-notification process serves a different purpose.
Keep the artist involved, check permissions and compare the result with the original before a track is released.

If you're producing a soca release, you have a deadline and a sound to get right. Will a new tool help? Perhaps. But if it takes an afternoon to repair the result, or leaves you unsure whether you can use the recording, the first quick export has not done the job. Start with the song in front of you. An industry forecast cannot make those choices for you.

What the forecast tells us

In a 6 July 2026 Antigua Observer report, communications consultant and creative-technology trainer Timothy Temal estimated that 60–80% of productions for Trinidad and Tobago's 2027 Carnival season would include AI in some form. The newspaper attributed the comments to its Observer AM interview. It did not provide a sampling method or name the artists and productions behind the estimate.

Temal also described making what he considered an early AI-assisted soca track in September 2024, and reported that initial resistance had given way to wider use. Those are his account and assessment. They do not establish the first use of AI in the genre or measure how audiences now respond to it. This article has no independent adoption survey to confirm the forecast.

The forecast is still a reason to ask practical questions. If a studio says it uses AI, does that mean help with an email, a vocal repair or a newly generated melody? Each affects the work differently. Counting all of them together would tell an artist little about who wrote the song or how it sounds.

For scheduling, Harts Carnival lists its 2027 road dates as Monday 8 and Tuesday 9 February. Those are not the limits of the whole Carnival season. Studios should work back from their own agreed delivery dates and the events they are supplying, checking the relevant organisers rather than relying on a generic week-long date range.

Decide what you want to improve

Start with one task that can be described and heard. Perhaps a rehearsal recording has a steady background hum, or a singer wants to compare two arrangements. “Make it professional” leaves too much unstated. Ask the artist what must stay: a phrase, a breath, the feel of the rhythm or the way the voice sits against the instruments.

Adrian Dunkley's People First, Problem Second, AI Last method offers a useful order for that conversation. Agree whose problem you are solving, define a satisfactory result, then consider the tools. It does not mean every production needs AI. A new take, a familiar editing tool or a clearer instruction to a musician may solve the problem with fewer complications.

The Audacity manual illustrates why the task matters. Its noise-reduction effect is designed for steady background noise, using a sample of that noise as a reference. The documentation warns that overly aggressive settings can damage wanted audio. This ordinary audio-processing example shows why you should listen for what was lost as well as what was removed. It tells us nothing about local AI adoption.

If a tool generates a new part, the review changes. Does the melody fit the artist's intention? Did it alter a word or flatten a deliberate rhythmic choice? Does something sound familiar enough to warrant a rights check? A clean-looking waveform will not answer those questions. Keep the artist and the person responsible for the recording in the decision.

Check permission before uploading

Having a client's file does not mean you may do anything you want with it. Before sending a client's vocal, an unreleased song or another writer's lyrics to a service, agree what is allowed. Explain which service receives the material and what the proposed task involves. A casual request to improve a mix should not be treated as blanket approval to train a voice model or publish a demonstration.

Read the provider's current terms. In the Suno terms reviewed on 13 October 2026, the content licence is broad and includes uses connected with service and model improvement. Its terms also set conditions for commercial use and state that the company does not guarantee copyright will exist in an output. A subscription receipt therefore cannot answer every question about a release.

Suno's copyright help page also distinguishes its arrangements with users from copyright eligibility, which can vary by jurisdiction. Its separate lyrics guidance tells users to obtain permission for someone else's lyrics or other content. That guidance cannot settle the rights to a Trinidadian artist's song. If the planned use is unclear, ask the provider and get qualified advice before release.

For a practice session, use material the team created and has agreed may be used for that purpose. Keep an untouched original and save the tool settings and outputs separately. Do not clone another person's voice or imply they performed on a track without their informed permission. The ability to produce a convincing imitation is not consent. For Suno’s specific Voice Model feature, its terms checked on 13 October 2026 allow only the account user’s own voice. They prohibit creating that model from another person’s voice, so the other person’s permission does not override this feature restriction.

Copyright and COTT are different questions

The Trinidad and Tobago Intellectual Property Office says protection for qualifying copyright works is automatic and does not require registration. Its guidance distinguishes authors' rights from related rights in performances and sound recordings. A song can therefore involve several people and interests. This general information does not decide whether a particular AI-assisted output qualifies or who owns it.

COTT's Notification of Work form asks for contributor names and royalty shares, with separate fields for performance and mechanical shares. Its instructions note that being a studio producer or arranger does not necessarily make someone the composer. They also call for appropriate authorisation when notifying arrangements, including sampling, of copyright works. The form includes an assignment of rights and signatures, so it should be read carefully rather than treated as routine proof of ownership.

Before completing a notification or signing an agreement, settle the actual contributions with the other people involved. Ask COTT how its current process applies to the work and have a qualified adviser review uncertainty about rights. Do not list a person as a writer simply because they ran a tool, or promise that submitting a form clears a disputed sample.

This article does not claim that every AI-assisted song is unprotected, that every generated output is safe to sell, or that training-data questions have one answer everywhere. You will need the facts about the work and the contracts to get useful advice. Start by recording who contributed what. List any outside material and the permissions you hold beside it.

Try one controlled comparison

You could try the following studio exercise. We have no measured productivity result for it. Choose a short passage that you have permission to process. Write down the problem before changing it. Keep a version made with your usual workflow, then make a second version using the proposed tool. Have the artist listen without being told which version used it, where that is practical.

Use a comparable listening level and ask about specific differences: the clarity of the words, unwanted sounds or changes to the feel. Keep the notes, including a preference for the original. If you prefer the result, try another passage before committing more work. One good attempt cannot tell you how the tool will handle your catalogue or the rest of the season.

Record the whole time spent. Setup, repeated attempts and fixing mistakes belong in the calculation alongside the first processing pass. Include subscriptions or usage charges when comparing costs. An export produced quickly can still be expensive if the engineer spends the rest of the afternoon making it usable.

There also needs to be a stopping point. If the revised file creates a rights question, changes the singer's meaning or cannot be reproduced reliably, pause that use. Return to the earlier version or another method. Keep a version you can release if the new approach fails. You do not want the deadline to arrive while a basic permission question is still unanswered.

Make the handover clear

Before sending a final master, the producer and artist can agree a short record of the work:

Identify the people who wrote, performed and produced the track. Confirm credits and agreed shares with them.
Record the tools used and the task each performed. Keep dated terms or licences relevant to any generated material.
Keep permission records for supplied lyrics, samples and recordings. Separate an internal reference from material cleared for release.
Check the distributor's current rules and any event or competition requirements, including disclosure where required.
Confirm who signs off the final version, where the original files are kept and who handles a later correction or rights query.

These notes need not become a public account of every edit. Their purpose is to let the people responsible answer a question later without reconstructing the session from memory. Any public description should be accurate about what the tool did. Do not claim that an artist sang a generated performance or that a demonstration proves endorsement.

The effect on livelihoods needs evidence too. This review has no measured change in Trinidad and Tobago's session work, engineering fees or artist income caused by AI. Studios can track their own hours and payments, and discuss changes with the musicians whose work is affected. Ask the people you work with how a proposed change would affect their pay and the work they want to do.

Before the track leaves your studio, can the artist explain and approve what changed? Can you account for the material in the master? Those are questions you can answer now. Temal's forecast will need later evidence; this recording needs your judgement today.

Frequently asked questions

Will 80% of Carnival 2027 music use AI?

The 60–80% range is Timothy Temal’s forecast, reported on 6 July 2026. This review has no independent survey confirming the share.

Does a paid AI subscription guarantee copyright?

No. Provider permissions and copyright eligibility are separate questions. Check the applicable terms, the actual contributions and local legal advice before commercial release.

Is COTT notification required to create copyright?

The Intellectual Property Office says copyright protection for qualifying works is automatic. COTT’s notification and rights-management process is distinct; read its terms and ask how they apply to your work.

What is a sensible first test?

Choose one passage you have permission to process, preserve the original and compare quality and full cost with your usual workflow. Let the artist take part in the decision.