What the tested workflow does
- Prepare permitted audio. Start with an MP3, WAV, or M4A file that you are allowed to use.
- Select the file in the converter. The browser decodes and resamples the audio locally.
- Run the note model. The page loads its model files from this same site and analyzes the audio in the tab.
- Review the draft. Adjust note sensitivity or minimum note length if needed.
- Download the MIDI. The page creates a
.midfile in the browser for you to save.
What we verified in a local browser
We tested the current page in headless Chromium with a generated four-tone WAV. The recording contains four one-second test tones and no copied video, title, channel name, or copyrighted performance.
| Checked item | Local result | Meaning |
|---|---|---|
| URL inputs on the converter | 0 | A video link cannot be pasted into the current interface. |
| Text file containing a test URL | Rejected | The button stayed disabled and the page requested MP3, WAV, or M4A. |
| Generated WAV | 5 note events | The controlled file produced a downloadable MIDI draft. |
| Conversion requests | 2 same-origin GET requests | The browser fetched the model manifest and model data from the local site. |
| POST or external requests during conversion | 0 | The tested audio conversion did not send a request to an upload endpoint or another origin. |
The five-note count is a result for this controlled file, not a general accuracy promise. Network observations describe the current code and this test run; browser extensions, developer tools, or future site changes can alter what appears in a user's network log.
Why some YouTube-to-MIDI tools accept links and this one does not
A link-based service must obtain audio from the video URL before transcription. This converter deliberately starts later in the workflow: it receives a file from your device and never includes a URL-fetch step. The difference is input handling, not a different MIDI file format.
Public competitor pages we reviewed describe both approaches. Some accept a YouTube URL and retrieve the audio for transcription, while others also allow a local audio upload. On this site, the supported path is local-file input only.
Does the audio leave the browser?
In the tested run, no POST request and no request to an external origin occurred after the WAV was selected and conversion began. The two conversion-time requests were GET requests for /basic-pitch-model/model.json and its model-data file on the same origin. The source audio itself was decoded from the chosen local file inside the page.
This is narrower than saying the page makes no network requests: the application code and transcription model still have to load. It means the current conversion path has no server-upload call for the selected audio.
What if you already have a permitted audio file?
Open the YouTube-to-MIDI file converter, select the file, and generate the draft. If the result contains short duplicate detections, follow the extra-note cleanup guide. If several instruments are mixed together, read the track-separation boundary before converting.