Terms of ServiceContact UsReport a Bug

© 2025 MIXTAPE. All rights reserved.

Guide: the Secrets to Better AI Songs

• Why incrementally adding to your songs is better • How NOT to create songs (if you want to own them) • Why mastering AI songs will give you an edge • What tech has the upper hand* • More…

In this guide, we will cover:

• Why incrementally adding to your songs is better
• How NOT to create songs (if you want to own them)
• Why mastering AI songs will give you an edge
• What tech has the upper hand*
• More…

Current tech stack:
Suno & BandLab

*AI tech is always evolving. To stay in the loop, consider subscribing to our Mixtape Newsletter!


What I use to create my AI music

As you already know, AI music generation is new, and constantly changing. This guide is current (Q2 2025) and lists the tools and methods I use for creating AI music.

Why I use Suno

If you have any plans on putting your music on streaming services and monetizing your music, you must first own it.

Not all AI music services provide ownership of the generated material. This is why you must review their terms of service. None of this information is legal advice, but simply my thoughts and methods.

Here’s a current overview for the top-tier music generators:

image.png

These may be the top choices for today, but no one knows what tomorrow brings. Join our community to stay updated on any changes. Disclaimer: none of what I cover in these guides constitutes legal advice.


Song Generation with Suno

There are many great generator services, but I use Suno for it's groovy sound and overall sense of musicality. It's very good, and can produce good results, right out the gate. However, there is a lot you can do to improve the output, to reliably produce radio-ready songs. If you want to have a dependable process that can churn out one great song after the other, we'll have to break down the creative process in some steps.

Why incrementally adding to your song is better

The method I use:

Write one verse. For example:
“Oh hold on detective
let’s look at the evidence
in case you suspect me
let’s get real objective”

Generate tracks until you hear something that stands out. Steer clear of:

  • Bad audio clarity

  • Overtones, crackling, shimmering

  • Muted sound (common in Suno v4.5)

The reason we look for those things is that if they make it into the song early, they will begin to corrupt the remaining sections and make them unpleasant to listen to.

Create till you hear something you really like. Be picky, and trust your intuition. I usually make about fifty to a hundred generations of a song, refining the lyrics along the way. From there, I usually end up liking 5-10 of these. After I have liked them, I choose my favorite out of those 5-10 songs.

Once you have found that one clean generation you like (the beginning to your song) use Suno’s Extend function to start adding to the end of the first verse.

Generate until you like the result, then choose Get Whole Song to fuse the first part and the extended addition into one.

Now keep adding to the song, extending then making it a whole song. Keep doing this until the song is done. Experiment with song structures, as the best way to learn is to try.

Example steps:

The ending of the song can be tricky, as the songs tend to want to go on indefinitely. Sometimes it helps to do steps 5 and 6 in the same generation, and to end the lyrics with [End] to make the track stop in time.

Which brings us to: how NOT to create songs (If you want to own them)


How NOT to create your music

In addition to your AI music service giving you rights under their terms of service, we must make sure that the songs we make qualify for copyright protection. Different countries have different standards, but one common theme exists:

To gain copyright, the songs generated must, among other things, show “substantial human involvement.”

Personally, I take the following steps:

  • I write my own lyrics (a common-sense argument for human involvement)

  • I am iterative and selective by creating the song in steps (involvement)

  • I document my process and involvement

*None of this information is legal advice, but simply my thoughts and methods.

AI music is a new area for copyright. If you want to stay updated you can subscribe to the Mixtape newsletter.

Once you have made a song that you both own and think is good enough for release, the next step is to make it even better.

Which brings us to: why mastering AI songs can give you an edge

Mastering Your Songs


This step is optional but advisable.

Mastering ensures your music sounds balanced and consistent across all platforms and devices, aligning your track with streaming services’ loudness standards.

In other words: mastering can give your song a professional, radio-ready feel.

There are many ways to go about this:

Having tried all of these, I have settled on the latter two options. However, these are not necessarily right for everyone. It depends on your budget and requirements.

Steps:

  1. Download your finished track from Suno as a .wav file. This is your “mix.”

  2. Use this mix when mastering with any of the above (or other) services.

You can get good results with BandLab, but to do so you might want to spend some time and generate many options to compare/choose between.

When selecting an AI master, pick one that sounds pleasant, punchy, and balanced with clear depth and width that keeps you engaged without ear fatigue.

Some generated tracks already have quite the balanced audio image, in which case you might choose to do minimal mastering. If you don't want to change the feel of the song at all, you can simply adjust the master volume of the track so that it is bumped up (increased) without being too loud (causing clipping). This will ensure that your track volume is not lower compared to others when it lands on streaming sites. Usually, the tool to do this is called a "Maximizer."

At the end of the mastering process, you’ll have a mastered track in the form of a new .wav file.

By mastering your track, you’re already ahead of the competition. Congratulations!

Next: Additional steps for quality AI songs


General Tips and Tricks

With music being subjective, there is no one-size-fits-all. Some people will gravitate towards certain sounds, and music services.

With that being said, I mainly use v4.5+ -- Suno's newest model.

Having created around thirty five thousand generations with Suno v.3 and its versions leading up to v4.5+, here are some observations:

  • Line breaks induce a long pause, whereas commas and periods make shorter pauses. You can break the middle of a verse line for intended effect.

  • If you want to pronounce words a certain way, you can play around phonetically (e.g., “destnee” instead of “destiny”).

  • A clear rhythmic structure produces more predictable results, i.e.: an even number of syllables; ensure the last words of the second and fourth lines rhyme (ABAB) or that the lines rhyme directly (AABB).

  • Model difference: v4 prefers short style prompts; v4.5 can handle longer prompts (if needed).

  • Be careful not to add too much text per generation: I use 4–8 lines max, and build on the song in increments.

  • Create a Suno Persona using one of your favorite tracks so you can make more similar songs. Set audio unfluence to 7-25% depending on desired results (I personally set it in the lower range)

  • [Brackets] are powerful for adding instruments and sounds; (parentheses) are good for adding backup vocal effects, whispers, echoes, etc.

  • When extending, do so from a moment with no vocals to prevent words from getting cut off or jumbled.

  • If you hear noise pollution, backtrack by going to an earlier version of the track, or extend from an earlier point.

  • Don’t be afraid to spend credits on a track you believe in. On any given song, I spend about 30% of my credits finding a good, catchy beginning, and then 70% to build it step by step.

Remember, what I list today might change tomorrow. These guides will adapt over time. As a free member, you will have instant access to these updates.

Perhaps more importantly, we have a community that, collectively, presides over more experience than one individual ever could. If you have a question or want to chime in, join the conversation on the Mixtape Community.


This concludes the Guide: Secrets to Better AI Songs. Thank you for reading and for being a supporter of Mixtape.

© 2025 Lost Mixtape. All rights reserved.

This guide and all related content are the intellectual property of Mixtape and its creators. You are granted a personal, non-transferable license to access and use these materials for your own informational purposes only. Redistribution, resale, reproduction, or sharing of any part of these materials with third parties is strictly prohibited.

Informational and Educational Purposes Only
The content provided in these guides is for informational and educational purposes only. Nothing in these materials constitutes legal, financial, or professional advice. Use of this information is at your own risk. For advice specific to your situation, consult a qualified professional.

No Professional-Client Relationship
Accessing or purchasing these guides does not establish any professional-client relationship between you and Lost Mixtape or its creators. No such relationship is created by your use of these materials.

No Warranties or Guarantees
While every effort is made to ensure the accuracy and currency of the information, Lost Mixtape or its creators make no warranties or guarantees, express or implied, regarding the completeness, accuracy, or reliability of the content. The tools, platforms, and methods described may change over time, and results may vary. No outcome or result is guaranteed.

Limitation of Liability
Lost Mixtape and its creators shall not be liable for any direct, indirect, incidental, consequential, or special damages arising out of or in connection with the use of these guides or any content therein, including but not limited to errors, omissions, or inaccuracies, or any loss or damage incurred as a result of reliance on the information provided.

Third-Party Tools and Services
References to third-party tools, platforms, or services (such as Suno, Midjourney, BandLab, etc.) are for informational purposes only and do not constitute an endorsement or affiliation. Users are responsible for reviewing and complying with the terms of service and licensing agreements of any third-party products or services mentioned. Lost Mixtape or its creators is not responsible for changes to third-party services or for any issues that may arise from their use.

Modification and Updates
Lost Mixtape and its creators reserves the right to update, modify, or discontinue any part of these guides or this legal section at any time, without prior notice. Continued access to the guides after changes constitutes acceptance of those changes.

Contact and DMCA
If you believe your copyright or other rights have been infringed by content in these guides, please contact us on our website for prompt resolution.

Sidebar