AI music prompting guide: Generating original tracks for your videos with MusicGen

You’ve spent hours perfecting your video, the visuals are crisp, the story flows beautifully, and then… You tag on a generic track that makes the whole thing feel like a corporate training video from 2003.

But finding the perfect track doesn’t need to be hard. Have you heard of MusicGen?

If you’re not acquainted, MusicGen is our AI-powered music generator that turns text prompts into actual, usable tracks. And it’s part of your Envato subscription at no extra cost!

Like any creative tool, there’s a difference between using it and using it well. The gap between “AI-generated background noise” and “wait, this actually works” comes down to knowing how to prompt properly.

Whether you’re entirely new to AI music generation or have been getting mediocre results and want to level up, this guide will walk you through everything you need to know to create original tracks that enhance your videos instead of just filling the silence.

While anyone who can type a couple of words into MusicGen can create quality music using text prompts and presets, how well the result fits your specific need depends entirely on how well you describe what you want. 

Which makes sense. Imagine if you were working with a real human composer. Saying to them, “Make me some happy music for my video,” won’t get you the same results as an in-depth, detailed instruction or brief about what you’re looking for.

Humans and machines need concrete descriptors, musical terminology, and clear emotional direction for best results.

Think of it like ordering coffee. “Something good” gets you whatever the barista feels like making. “Double shot cortado with oat milk and a hint of vanilla” gets you exactly what you want. The same principle applies here.

Now that you know why good prompting matters, let’s get practical.

Let’s walk through the actual process, step by step.

Head to envato.rseotools.com/__labs__/music-gen. You’ll land on a clean interface with a text box for your prompt and some preset options to get you started.

You’ve got two ways to prompt, but they can be used in combination:

For your first track, try a preset to get a feel for how the system works. Once you’ve dipped your toes in, let’s get to why we’re here — to master prompting for your video, so let’s dive into custom text generation.

Here’s where the magic happens (or doesn’t, depending on your prompt). Let’s start with a basic example:

Simple prompt: “upbeat music” 

Advanced prompt: “energetic indie pop track with driving drums, bright acoustic guitar, and uplifting melody, 120 BPM, perfect for lifestyle content”

Notice the difference? The good prompt includes:

Pro tip for lyrics: If you’ve toggled on the lyrics feature, you can either let MusicGen generate vocals automatically based on your mood and genre, or provide your own lyrics. Try adding phrases like “with female vocals about summer adventures” or include actual lyrics in quotes: “Chorus lyrics: ‘We’re chasing the sunset, living our best life'” (or something much less corny!). The AI understands song structure, so you can specify verses, choruses, and bridges.

Hit generate and wait about 30-60 seconds. You’ll get multiple variations of the same prompt, each with different lengths and interpretations. Listen to each one — the waveforms give you a visual preview of the energy and dynamics.

Found one you like but want to tweak it? Don’t just regenerate the same prompt — refine it. Maybe the energy is right but you need different instruments, or the mood is perfect but the tempo is off.

An evolution of the original prompt could look like:

See how each version keeps the core elements that worked while adjusting the specific aspect that didn’t quite hit the mark?

Love what you got? Use the extend feature to make it longer, or download it as-is. Each generation comes with a license file for commercial use.

After generating thousands of tracks (some tragic, some surprisingly good), we’ve figured out what separates the wheat from the chaff. Great prompts have a structure that gives the AI exactly what it needs to work its magic.

Here’s how to structure yours:

Be specific. “Electronic” could mean anything from ambient drone to hardcore techno. Better options:

Name the instruments you want to hear:

Give the AI rhythmic guidance:

This is where you paint the emotional picture:

Help MusicGen understand the purpose:

Guide the vocal style and lyrical content:

Let’s break down some real prompts that consistently produce good results:

“Upbeat acoustic pop track with finger-picked guitar, subtle percussion, and warm vocals, 110 BPM, bright and optimistic mood, perfect for lifestyle and GRWM content”

Why it works: Specific genre, clear instrumentation, defined tempo, mood descriptors, and use case.

“Cinematic orchestral piece with string section, gentle piano, and subtle brass, building from quiet contemplation to emotional climax, 85 BPM, evocative and moving”

Why it works: Sets cinematic expectations, describes the emotional journey, includes dynamic changes.

Need inspiration for a slow-build emotional arc? Dive into our breakdown of Hans Zimmer’s Interstellar soundtrack—a masterclass in evolving orchestration you can echo in your prompts.

“Modern corporate track with clean electric guitar, steady drums, and subtle synth pads, professional and confident feel, 100 BPM, motivational without being cheesy”

Why it works: Addresses the specific challenge of corporate music (avoiding the “cheesy” trap corporate music can veer towards), clear energy level.

Learning from bad examples is just as valuable:

“Something good for my video” Problem: No genre, mood, or context. The AI has no direction.

“Relaxing heavy metal with soft drums” Problem: Conflicting descriptors confuse the system.

“Rock pop jazz classical electronic hip hop fusion with every instrument” Problem: Too many genres and elements create musical chaos.

“I need a track that sounds like driving through the countryside on a summer evening while thinking about your childhood and feeling nostalgic but also hopeful about the future with birds chirping and…” Problem: Too much narrative, not enough musical direction.

Once you’ve mastered basic prompting, try these advanced approaches:

Combine genres strategically: “Indie folk with electronic elements” or “Classical orchestration with modern hip hop beats.”

Describe how the track should evolve: “Starting melancholic, building to hopeful climax.”

Use musical terms instead of artist names: “Minimalist piano style” instead of “like Ludovico Einaudi.”

Include production descriptors: “Warm analog sound,” “Crisp digital production,” or “Lo-fi aesthetic with vinyl crackle.”

When lyrics are enabled, get sophisticated with vocal arrangements: “Lead vocal with subtle harmonies in the chorus,” “Call and response between two vocalists,” or “Ethereal backing vocals creating atmosphere.”

If you’re shaping both sound and visuals, pair this guide with our companion piece on AI video prompts, which shows you how to script footage that syncs perfectly with your custom soundtrack.

Different video types need different approaches:

Focus on immediate impact and loop-ability:

Emphasize variation and development:

Balance engagement with professionalism:

Prioritize mood support over distraction:

Issue: Track sounds too “AI-generated” 

Solution: Add human elements to your prompt like “live drums,” “analog warmth,” or “slight imperfections.”

Issue: Wrong energy level 

Solution: Be more specific about tempo and dynamics. “Energetic” can mean different things.

Issue: Doesn’t fit your video 

Solution: Include your video type in the prompt. “Documentary background music” yields different results than “social media track.”

Issue: Too repetitive 

Solution: Ask for “dynamic arrangement” or “multiple sections” in your prompt.

Issue: Vocals don’t match the song mood

Solution: Be more specific about vocal delivery. Instead of just “happy song,” try “bright, energetic vocals with a smile in the voice” or “playful, almost laughing delivery.”

Issue: Lyrics feel generic

Solution: Provide your own lyrical themes or specific lines. “Lyrics about small-town nostalgia” is better than “nostalgic lyrics.” Best results come from providing actual verse/chorus lyrics.

The key to getting consistently good results is iteration. Don’t worry if you don’t nail it first try. Professional music producers often go through dozens of iterations (and that’s on a good day) before they land on the one. Same applies here.

Start with a basic prompt, generate a track, then refine based on what you hear. Maybe the genre is right, but you need different energy. Maybe the mood is perfect, but the instrumentation is off. Each generation teaches you more about how to communicate with the AI.

Most importantly, remember that MusicGen is just one part of your creative toolkit. Combine it with ‘Sounds Like’ Search to find reference tracks, use Claim Clear to avoid copyright issues, and explore our broader music library when AI generation isn’t the right fit. With over 300,000 tracks, including PremiumBeat’s premium catalog, you’ve got options whether you need AI-generated originals or professionally produced tracks.

If you’re using MusicGen regularly or browsing our full catalog for licensed music, a music subscription is the easiest way to unlock everything — without constantly worrying about usage rights or extra costs.

The goal here is to give you a variety of tools to create exactly what you need for your project, when you need it.

Great prompting is part art, part science, and part persistence. The examples and principles in this guide will get you started, but as with all things worth doing, the best way to improve is by experimenting.

Start with prompts that match your content type, then modify them based on your specific needs. Pay attention to what works and what doesn’t, and build your own library of effective prompts over time.

So, next time you’re stressed about finding a track for your video, take a breath and remember: The perfect track might be one prompt away. Feels nice, doesn’t it?