AI transparency

Every track here is AI-generated

No exceptions. Each track in the library is produced by a music generation model, then listened to by me, selected, loudness-normalised and tagged.

I say this in four places, deliberately repeated:

• On every track page, directly under the play control rather than buried in the footer.
• In the metadata of every downloaded file, so the information travels with the file once it leaves this site.
• In the description of every video on my channel.
• In the synthetic content disclosure flag I set on every video.

Why I am explicit about it

Three independent reasons, each sufficient on its own.

First, the model provider's terms. They forbid me from stating, implying or presenting output in a way that suggests a particular person composed it.

Second, Vietnamese law on labelling AI-generated content.

Third, the video platform's policy requiring disclosure of synthetic content. The platform states plainly that the label by itself does not reduce recommendation or monetisation, so there is no reason to dodge it.

There is a fourth reason that is nobody's requirement: if I hid this, everything else I say would deserve doubt.

What this means for you

In many jurisdictions, including the United States and Vietnam, a work generated entirely by artificial intelligence is currently not protected by copyright.

That cuts two ways.

In your favour: usage is very open, because there is no author's rights holder standing behind it to restrict you.

Against you: I cannot grant you exclusivity over a track, and I cannot stop anyone else from using the same one. The licence I grant is a contract between you and me, not a claim of authorship.

If you need exclusive music, what you need is a commission, not a free library. Saying that plainly is better than letting you find out later.

What I do not do

• I do not invent artist names for these tracks.
• I do not attach a biography or a photograph of a person who does not exist.
• I do not register these tracks with rights-identification systems, and I forbid users from doing so.
• I do not claim my music is entirely safe from claims. No library anywhere can promise that.

What I check before publishing

An AI-generated track can still reproduce the melody of an existing work. That risk is real and I handle it in three steps.

Listening comparison. Anything that sounds familiar is dropped outright, no deliberation.

Technical measurement. I check loudness, true peak, unexpected silences and signs of an abrupt cut.

A check against the video platform's identification system before release. I upload a private test video and wait, letting it compare against the largest reference database in existence. Anything that draws a claim is dropped permanently.

These three steps do not reduce the risk to zero, but they remove most of it, and I apply them to every track rather than only to suspicious ones.