One clip, heard silent, then with music, effects and both — and what “royalty-free” doesn't cover.
The full script, word for word — 612 words, about 3 minutes to read. Current as at August 2026. AI tools change quickly; if something looks different when you try it, check the product's own help pages.
Hi, I'm Mel. Sound can change the meaning of a shot before the picture changes at all. We took one eight-second lion clip and made four versions: silent, music only, sound effects only, then both together. Listen first. The licence question comes straight after.
The picture is identical each time. The original purr has been removed from these comparison copies so it cannot favour one version. The generated music, wind, birds, footsteps and breathing were made for this test. Listen for what each layer asks you to feel. This comparison does not rank the four versions.
Silence makes the lion observational. Music gives the shot an argument: majestic, tense, playful or sad, depending on the cue. Effects make the frame feel physical. Grass, breath and distant birds put the animal in a place. Together, music carries interpretation while effects carry presence. If the two compete, the result feels busy. Choosing the source is only half the job. The other half is permission.
There are four useful routes. Record on location when the place or action needs to be heard. Use a documented free library when the destination and licence are clear. Generate bespoke audio when you need a particular mood or effect. Or license a track from a dedicated library when client work and repeat use justify it. Music inside an editor is convenient, but convenience does not replace reading its terms.
A track available inside one platform is not automatically cleared everywhere else. YouTube describes its Audio Library as copyright-safe on YouTube, then says it cannot advise on off-platform problems. Canva has separate conditions for Pro Music and Popular Music. If the video moves to another channel, a client, an advertisement or a paid product, check the licence again. Save the track name, source, download date and licence evidence with the project.
Royalty-free does not mean free of charge, free of conditions or free of copyright. It usually means the asset is licensed once without a new payment for every permitted use. The licence still defines the media, audience, territory and restrictions. “Copyright-free” and “I found it online” are not production standards. The permission controls the use.
AI-generated music needs the same discipline. ElevenLabs says free-plan output is non-commercial and requires attribution when shared. It says content generated during a paid subscription may be used commercially, subject to its terms and the rights you hold. That timing changes the licence. Upgrading later does not rewrite the licence on an earlier generation. We made these test sounds after the paid plan began, and kept the source files as evidence.
A useful sound prompt describes the source, distance, environment and what must be absent. “Dry grass footsteps, close and natural, no roar, music or voices” is more controllable than “lion sounds”. For ambience, ask for steady wind, sparse distant birds and no sudden calls. Prompt influence can improve adherence, but higher is not automatically better. Generate options, reject distractions and keep the quietest useful layer.
For a quick internal video, an editor library may be enough. For YouTube, its Audio Library gives a documented starting point. For bespoke sound, generation earns its place when the plan covers the intended use. For client, advertising or repeated commercial work, a dedicated licensed library can buy certainty and an audit trail. In every route: confirm permission, save evidence, then make the sound serve the picture.
Start with the silent cut. Add effects to establish place, then music only if it improves the meaning. Check the licence before the edit becomes expensive, and keep the evidence beside the files. In the next video, we will take these layers into the edit and balance them around the voice.