AI Music Generation: How to Make a Track and What It Costs

Pricing is per minute, and that is the main thing to understand before the first generation. MiniMax Music has two generations: v2 is fast and cheaper, v2.6 is more careful with the arrangement and vocals. Start with v2 to catch the style, and build the final on v2.6.

Generating music with AI in 2026 is no longer an experiment but a working tool: in a couple of minutes you get a track with vocals and an arrangement from a text description. This guide covers how to make music with AI in Pixyn with the ElevenLabs Music and MiniMax Music models: describe the style, add your own lyrics, get background music for videos and Reels. Everything happens in the browser. We show the step-by-step process and give ready prompts.

TL;DR

• What it does. ElevenLabs Music and MiniMax Music in Pixyn generate a track (vocals plus instrumental) from a text prompt: you describe the style and can add your own lyrics. Genres range from pop and rap to rock, lo-fi and ambient.
• How you control it. The genre, the mood, the type of vocals and the language are set in words in the prompt. MiniMax Music has a separate field for the lyrics, and ElevenLabs Music and MiniMax v2.6 have an instrumental switch. ElevenLabs Music makes tracks up to 5 minutes long.
• Where to make it. The web Studio, the Telegram Mini App and the visual Workflow for batch generation.
• What it is really for. Backgrounds for videos and Reels, jingles, demo sketches, podcast intro music, a track "for the mood".
• What it costs. Pricing is per minute: MiniMax Music v2 from 10 tokens, v2.6 from 30, ElevenLabs Music from 40. The base includes one minute, and each next minute costs the same. A three-minute track on v2 is three minutes at the price of a minute, not one price "per track".

What AI music generation is and where it helps

An AI for music works much like an AI for pictures: you give a text task (a prompt), and the model returns a result, only it is an audio track rather than an image. ElevenLabs Music and MiniMax Music can compose a melody, an arrangement and sing vocals at the same time. So the output is not a backing track but a finished song with words.

Where it really comes in handy:

• Social media content. Background music for Reels, Shorts and TikTok, so you do not fight other people's copyrighted tracks and takedowns.
• Marketing and advertising. A brand jingle, a music bed for a product promo video.
• Podcasts and YouTube videos. An intro, transitions, a background track under a voice-over.
• Personal projects. A birthday song with a name, a track for fun, a demo sketch for a musician to hear an idea quickly.

The frame to understand right away: AI music is about speed and drafts. For a video background and a demo it is excellent; for a release under your own name it is a starting point that is usually refined.

How to generate music with AI in Pixyn: step by step

Below is the basic "from description to track" scenario. The whole process runs in the browser.

1. Sign up. Open the sign-up page: it takes 30 seconds via email, Telegram or Google. New accounts get welcome tokens to try it without paying.
2. Go to the Studio. Open the Studio, the main workplace for generation.
3. Choose a music model. In the audio category of the model list find ElevenLabs Music, MiniMax Music v2 or v2.6. The Studio brings up the fields needed for music generation itself.
4. Describe the track in the prompt. In the description field set the style and the mood in your own words: genre, tempo, instruments, emotion, what the track is for. The more specific, the more accurate the result.
5. Add your own lyrics (if needed). Want it to sing exactly your text? In MiniMax Music put the words in the lyrics field, marking the verses and the chorus (see the example below); in ElevenLabs Music add them to the prompt. No lyrics, and the model writes the words for the description itself.
6. Describe the vocals and the language in words. Write in the prompt "female vocals", "male vocals" or "lyrics in English". For a track without vocals, turn on the instrumental switch. For ElevenLabs Music you can also set the track length (up to 5 minutes).
7. Run the generation. Press the run button. A track is usually ready in about 1 or 2 minutes.
8. Listen and download. Open the result in the player, listen to it in full, and download the version you like. If it did not land, change the prompt and generate again.

A tip: if the vocals "drift" or the genre is off on the first try, do not rewrite everything; change one parameter at a time (first the genre, then the tempo, then the vocals). That way you find the sound you need faster and save tokens.

How to write a music prompt: structure and a ready example

A good music prompt is assembled from clear blocks. You do not need "literature": you need specific musical parameters.

What to include in the description:

• Genre and subgenre: pop, hip-hop, lo-fi, synthwave, rock, ambient, phonk.
• Mood: energetic, melancholic, calm, aggressive, dreamy.
• Tempo: slow / medium / fast (or an approximate BPM if you know it).
• Instruments: guitar, piano, synth, 808 bass, strings, beats.
• Vocals: male / female / duet / no vocals (instrumental), in words in the prompt.
• Purpose: "background for a video", "a catchy chorus", "a podcast intro".

A ready prompt (to copy):

Genre: dreamy lo-fi hip-hop, calm, medium tempo ~80 BPM.
Instruments: warm electric piano, soft 808 bass, light vinyl crackle, unobtrusive percussion.
Vocals: female, soft, in the background (or instrumental with no vocals).
Mood: cozy, evening, slightly melancholic, background music for videos and work.
Structure: a smooth intro, an even beat without sharp transitions, no loud drop.

A ready prompt with your own lyrics:

Style: pop ballad, piano + strings, medium tempo, female vocals, emotional, lyrics in English.

[Verse]
The evening settles on the city,
The streetlights wake and start to glow.

[Chorus]
And here we are again in the rain,
This moment will not let us go.

A lyrics tip: write in simple, phonetically "easy" words without heavy consonant clusters, so the vocals sound cleaner. Long words are what the model most often blurs.

Music for videos and backgrounds: a practical scenario

The most common request is background music for a video that does not infringe on someone else's copyright and does not get blocked on YouTube or Reels. Here AI music really helps.

How to make a background track for a video:

1. Decide on the length and mood of the video (an upbeat promo / a calm vlog / dramatic storytelling).
2. In the prompt, be sure to state the purpose: "background music for video", no vocals (if a narrator's voice will go on top), an even structure without sharp drops.
3. Generate 2 or 3 variants and choose the one that does not pull attention to itself.
4. If the track is shorter than needed, ask for a more "loopable" sound (loop-friendly): it is easier to edit end to end.

If there are many videos and you need a series of uniform beds, look at the visual Workflow in Pixyn: you can build a pipeline once and run a batch of generations in one click without repeating the settings by hand each time. See the Workflow page.

What it costs and why the bill goes by the minute

This is the one place where it is easy to be off five times over, so let's go through it separately.

Model	Included in the base	Each next minute
MiniMax Music v2	from 10 tokens, 1 minute	the same
MiniMax Music v2.6	from 30 tokens, 1 minute	the same
ElevenLabs Music	from 40 tokens, 1 minute	the same

So a three-minute track costs three times as much as a one-minute one, and a five-minute one on ElevenLabs five times as much. This is not a hidden surcharge: you set the length, and the price on the button is recalculated before the run.

The practical takeaway: look for the style on short excerpts. A minute is enough to hear whether the model hit the genre, the vocals and the mood. Once you have found it, generate the full length with the same prompt. Trying five one-minute variants on v2 costs less than one five-minute take.

How spending works in general: the card about tokens; current packs are on the pricing page.

FAQ

Can I generate music with AI for free? New Pixyn accounts get welcome tokens, enough to try music generation without paying. After that, it is by plan, on the pricing page.

Which music models does Pixyn have? Two music models are in the catalog now: ElevenLabs Music and MiniMax Music (v2 and v2.6). Both work from a text prompt: you describe the style and add lyrics if you want.

Can I add my own words so the model sings exactly them? Yes. In MiniMax Music put your text into the lyrics field, marking the verses and the chorus; in ElevenLabs Music add it to the prompt. The model performs exactly that text. Write in simple words for cleaner vocals.

Why was I charged more for a five-minute track than for a one-minute one? Because pricing is per minute: the base includes one minute, and each next minute costs the same. The number on the button is recalculated when you change the length, so check it before the run.

How long does generating one track take? Usually about 1 or 2 minutes per variant. The track itself can be up to 5 minutes long on ElevenLabs Music.

Can I make a track without vocals, just an instrumental? Yes. Turn on the instrumental switch (ElevenLabs Music and MiniMax v2.6) or write "instrumental, no vocals" in the prompt. It is handy for background music under a voice-over.

Ready to make your own track?

You can generate music with AI right now, and new accounts get welcome tokens.

Create your first track → Sign-up takes 30 seconds via email, Telegram or Google. Then open the Studio, choose ElevenLabs Music or MiniMax Music and describe the track in your own words.

For batch processing of dozens of tracks, see the Workflow page.

Questions about models and plans: write to support at @pixyn_support.