The sound on a short-form video rarely originates with the person who posted it. It is borrowed — from a platform library, from another creator’s clip, or from a track that has been passed along so many times that nobody involved could tell you where it started.

That borrowing has two consequences worth understanding. One is legal and affects creators. The other is acoustic and affects everyone listening, because each hop is another encode.

Two libraries, not one

TikTok, as the clearest example, operates two separate music catalogues, and which one you see depends on your account type rather than on what you are making.

	General library	Commercial Music Library (CML)
Who sees it	Personal accounts	Business accounts
Contents	Popular commercial music	Over 1 million pre-cleared tracks and sound effects
Cost	Free in-app	Free to any Business Account, no per-track fee
Cleared for ads?	No	Yes — organic brand posts and paid ads
Cleared off-platform?	No	No

A business account sees only the CML when choosing audio in the app, for anything promoting a product, service, business or sponsor — and that includes organic posts with no paid promotion behind them.

The rule that catches people is the one about intent rather than account type: if a creator on a personal account uses the general library for content that promotes a product, service or sponsor, the post is still subject to commercial music rules. The account label does not override the content’s purpose.

The licence stops at the platform edge

This is the limitation that causes the most expensive mistakes.

The CML licence covers TikTok content only. The moment the same video is repurposed to YouTube, Instagram or anywhere else, those tracks are no longer licensed for that placement.

Which is how a perfectly compliant video becomes a claim on another platform a week later: nothing about the file changed, only where it was posted.

For anyone cross-posting — which is nearly everyone — the practical implication is that platform-native libraries are a trap for reusable content. If a clip is going to live in more than one place, the audio needs to be licensed independently from the start.

What reuse does to the audio

Now the part that concerns this site.

When a creator uses an existing sound from the platform, they are not receiving the master. They are receiving the platform’s rendition of it — which, as the encoding chain piece covers, is AAC-LC at around 128 kbps.

That rendition then goes into a new video, which is exported, uploaded, and transcoded again.

Follow a popular sound through a few hops and the arithmetic is unforgiving:

Original master
  → platform encode          (generation 1)
  → creator A reuses it, exports, uploads
  → platform encode          (generation 2)
  → creator B reuses A's version, exports, uploads
  → platform encode          (generation 3)
  → your Bluetooth link      (generation 4)

Each encoder works from the damage the last one left, not from the original. This is why a sound that was crisp when it first circulated arrives weeks later sounding hollow and smeared — the audible signature of generational loss, most obvious on cymbals, sibilance and reverb tails.

It is also why the same track can sound noticeably different on two videos posted the same day. You are not hearing two masters. You are hearing two different points in a chain of copies.

What creators can do about it

Four things, in order of effect:

1. Go back to the source. If a sound exists as a purchasable or licensable file, use that file rather than lifting the platform’s rendition of it. This removes every accumulated generation at once.
2. Never re-upload a downloaded platform file. It is already at least one generation deep and often two. Re-uploading guarantees a visibly worse result than the clip you were copying.
3. Export high, from the cleanest source. Guidance for uploads sits at AAC 192–320 kbps, 44.1 or 48 kHz. You will not keep that bitrate — the platform re-encodes regardless — but you hand its encoder clean material, and that is what determines the output quality.
4. Keep the original audio separate. If the video may be cross-posted, mix with an independently licensed track and keep the project file. Retrofitting a licence to a finished, published video is far harder than planning for it.

What listeners can do about it

Considerably less, which is the honest answer.

You cannot undo generational loss. What you can do is stop adding to it and stop revealing it:

• Use a wire where practical, removing the final encode.
• Get the seal right — the fit guide — because thin bass makes generational damage sound far worse than it is.
• Choose a forgiving earphone for this material specifically, for the reasons set out in the piece on earphones for video.

The pattern underneath

Short-form platforms are built on reuse. The whole format depends on a sound circulating, being adopted, and being adopted again — and that is a genuinely good creative mechanic.

But the technical system underneath treats every reuse as a fresh upload, and every fresh upload as a fresh encode. The culture is designed around copying; the pipeline is designed as though every file arrives new.

The gap between those two facts is what you are hearing.