Suno v5.5 API Guide: AI Song Generation & Voices

You can turn a text prompt into a song in about 30–90 seconds, then move that file into review, storage, and release with a simple API flow.

If I had to boil this guide down, it comes to this:

• I set up API access with a key, headers, and polling

• I choose the right mode: description, custom lyrics, or music-only

• I shape the output with short prompts, voice cues, and clip extension

• I move drafts through stems, mixing, export, and approval

• I store files fast because result links may expire in 24–72 hours

• I check commercial rights, consent records, and approval logs before release

This guide is mostly for people who want to make song generation part of a working content pipeline, not just test it by hand. That includes app teams, ad teams, podcast producers, game studios, and film editors.

A few numbers matter right away:

• Tracks can run up to 240 seconds

• First poll should wait about 5–10 seconds

• Polling usually continues every 5–10 seconds

• Rate-limit errors like 429 need a short retry delay

• Dates for approval logs should be stored in U.S. format, like 07/16/2026

What I like about this setup is that the path is simple: send request → get task ID → poll status → download audio → store it → review rights.

Here’s the short version of the workflow:

In other words: this is less about “make a song” and more about building a repeatable music production system with clear inputs, review steps, and file handling.

Set Up Suno v5.5 API Access for a Working Project

Your first request comes down to three things: credentials, headers, and async polling.

API Keys, Authentication, and Base Request Structure

Create your account, then generate an API key in the dashboard's key management area. Before you move to production, set quotas and IP whitelists.

Never hardcode your key in source files. Put it in an environment variable instead:

SUNO_API_KEY=your_key_here

For local work, use a .env file. In production, use a secrets manager like GitHub Secrets or AWS Secrets Manager. On every request, send these headers:

• Authorization: Bearer YOUR_API_KEY

• Content-Type: application/json

The request flow is pretty simple. Send the POST request, store the task ID, wait 5–10 seconds, then poll every 5–10 seconds until the job returns SUCCESS and gives you an audio URL. ^[2][7][3] In most cases, generation takes 30–90 seconds. ^[2][3]

As soon as the track is ready, download it or move it into permanent storage, such as an S3 bucket. That matters because a temporary result URL may not stay available for long.

For the model field in your JSON payload, explicitly set "suno-v5.5" or the latest equivalent identifier your provider documents. If you need a certain model version, don't rely on defaults.

Using APIMart as the Workflow Layer

APIMart sends Suno requests through one server-side workflow, which keeps credentials out of client code. In plain terms, raw tokens stay off the client, and every call goes through the same integration layer.

In each request, set the model parameter explicitly so the request goes to Suno v5.5, or to whichever model you need for the next step. With access set up, the next piece is prompt design and output settings.

Generate Songs with the Right Prompt, Mode, and Output Settings

Once access is set up, pick the generation mode and output format before you write the prompt. In Suno API workflows, custom_mode and instrumental are the main fields to decide upfront before generating audio ^[5][8]. After that, you can shape voices and review steps to turn rough drafts into assets you can actually use, or integrate them into AI video generation workflows.

Description Mode vs. Custom Lyrics Mode

Description Mode (custom_mode: false) lets the AI write the lyrics and pick the musical style from a short text prompt. It's best when you want fast output with light direction. Custom Lyrics Mode (custom_mode: true) puts you in charge. You supply the exact lyrics, section tags like [Verse], [Chorus], [Bridge], and [Outro], plus style details such as genre, mood, and tempo. That makes it a better fit for ads, branded songs, and creator tracks with a set structure ^[5][3].

If you want instrumental-only output, set instrumental: true. Then use negative_tags to block vocals you don't want, such as "vocals, singing, voice" ^[6][5].

Prompt Patterns for Ads, Brand Music, and Creator Content

A simple prompt pattern works well here: [genre] [mood] [instruments] [voice] [era/influence]. Try to keep it to about 8–12 key terms so the model doesn't drift ^[3]. Too many mixed signals can throw it off.

Mode should match the job. Use description mode when you're aiming for atmosphere. Use custom lyrics mode when the song needs to carry a message, brand line, or story. You can also call out vocal texture directly with phrases like "soft female vocals", "deep male baritone", or "ethereal" to keep the sound lined up with a brand tone or channel style ^[3][5].

Output Planning for Production Use

Use the extend API to chain clips into a more unified 2–4 minute song. It also helps to include a dedicated [Outro] or some clear ending marker so the track wraps up cleanly ^[3].

For production work, generate more than one version, download each one right away, and save an instrumental cut alongside any vocal track. That gives you a fallback if a vocal version doesn't fit the ad, edit, or final review. It also helps to store every asset with its prompt and metadata in one consistent project folder.

Next, refine voices and edit steps so the best version is ready for review.

Handle Voices, Editing Steps, and Commercial Workflow Controls

Default Voices, Cloned Voices, and Personalized Voice Workflows

Suno v5.5 gives you three vocal workflows, and choosing the right one at the start can save a lot of cleanup later.

Default voices are the fastest path. You describe the vocal style in your prompt - something like soft female vocals, ethereal or deep male baritone - and the model takes it from there ^[3]. This option works well for fast commercial drafts when you don't need one fixed voice identity ^[5].

Cloned voices take more setup, but they give you a steady, recognizable sound. Use cloned voices only when you have clear, documented consent from the voice owner ^[4].

Personalized voice workflows push things further. These use Suno v5.5 edit flows and features like Swap Vocals or Swap Sound to fine-tune the final asset over several passes ^[1][5].

Once you've chosen the voice path, the next step is simple: move the draft through editing and sign-off.

From Draft Song to Usable Asset

The first generation is a draft. It's not the finished asset.

The workflow looks like this: Draft → extend → split stems → mix → export.

Start by checking the lyrics and vocal tone against your brief. Then split the stems so vocals and instrumentals sit on separate layers. That gives your audio team room to work inside a DAW like FL Studio or Ableton, where they can apply EQ, compression, and reverb. After that, export the final version as WAV and attach metadata like BPM, genre, and style tags before saving it to the project folder.

Download approved audio right away.

Licensing, Commercial Use, and Internal Approval Checks

Before release, clear the rights and log approval for every finished track.

First, confirm that the account used to generate the song includes commercial-use rights.

Then set up a light internal approval step. Flag any prompt or lyric set that mentions specific artist names or unauthorized likenesses. Requests tied to unauthorized likenesses or artist names may be rejected ^[5][6]. If you're working with cloned voice assets, keep the consent records with the audio file.

For audit tracking, log the Task ID, song title, duration, style tags, and approval date for each finalized asset. Use U.S. date format - like 07/16/2026 - and record who approved it ^[5][3].

Build an End-to-End APIMart Music Pipeline

A Simple Production Pipeline for Teams

Once your prompt, voice, and licensing choices are locked in, tie them together in one repeatable flow: brief in, approved audio out, then stored for video, ads, or social.

Send the brief to Suno through APIMart. The API returns a task ID, and your orchestration layer - like n8n - can run a polling loop until the job comes back as SUCCESS ^[2][5]. Hold off for 10 seconds before the first poll so you don't send extra requests for no reason.

The goal is simple: map each handoff to a tool or API so the process runs without someone chasing files or status updates by hand.

One detail matters here: generated audio URLs are temporary and can expire within 24 to 72 hours ^[5]. So your pipeline should download approved files right away and move them into permanent storage before that window closes.

After the basic flow is working, add failure routing so rejected generations go back to the prompt stage. The same goes for content_policy_violation responses - send them back for prompt changes so the queue keeps moving instead of getting stuck ^[5].

First Steps

Start with one workflow you can repeat, then grow from there only after it works end to end. Pick one use case first - an ad jingle, a branded background track, or a creator soundtrack.

Before anything goes live, check that your plan clearly includes commercial use rights ^[1]. Then use APIMart as the layer that connects music generation to your larger AI content workflow, so one song request can move through a system your team can use again and again.

FAQs

How do I choose between description, custom lyrics, and instrumental mode?

Choose based on how much control you want.

• Use Description mode if you want the AI to make the lyrics and composition from a text prompt.

• Use Custom Mode if you want to supply your own lyrics and song structure.

• Use Instrumental mode if you want a track without vocals, like background music or jingles.

What should I store with each generated track for commercial approval?

Keep a clear record for every track you make. At a minimum, store:

• the unique task ID

• the final generated audio file

• metadata such as song title, duration, and style tags or prompts

• the model version and creation date

That record makes it much easier to verify licensing compliance and keep an audit trail for commercial use.

How can I keep temporary audio links from expiring in production?

Audio links from the Suno v5.5 API are temporary. In most cases, they expire within 24 to 72 hours, so they’re not a good fit for long-term production assets.

Here’s the safer move: as soon as your status polling shows that the task finished successfully, download the audio file right away and send it to persistent storage.

That can be something like:

• Google Drive

• Your local file system

If you wait too long, the link may expire and the file could be gone when your app tries to fetch it.