Upscaled MP3: how to tell a file isn’t what it claims, and why it matters
Upscaling is a file in a high-quality format made from a low-quality one. Most often it’s a WAV converted from an MP3, or a 320 kbps MP3 re-saved from a 128 kbps one. The extension and the size say “high quality”, but the sound stays the same — with every loss of the source. Let’s see how to spot it and why it’s worth watching for before you send a track.
In short
- Converting an MP3 to WAV doesn’t bring back the lost data — only the file size changes.
- The main sign of upscaling is a sharp spectrum cutoff at 16–20 kHz in a file labelled lossless or 320 kbps.
- You can check a file on a spectrogram: in free apps or in the track analyzer.
- There are false alarms: lo-fi, vintage synths, deliberate filtering of the top end.
- Send an export straight from the project and ask your beatmaker for a WAV, not an MP3.
Contents
What upscaling is and where it comes from
MP3, AAC and other lossy codecs save space by throwing away what people hear worst: the very top of the spectrum, quiet sounds next to loud ones, part of the stereo information. What’s thrown away isn’t stored anywhere, so when you convert back to WAV there’s nowhere to get it from. You end up with an uncompressed file with MP3 content.
Typical reasons an artist ends up with an upscale:
- A beat was bought or downloaded as an MP3, vocals were recorded on top, and the whole track was exported to WAV.
- A sample or a reference was taken from YouTube or social media and dropped into the project as is.
- The track was once saved as an MP3 to send to a friend, and then the source got lost.
- A platform or a streamer asked for a WAV, and the MP3 was simply re-saved with a different extension.
What upscaling looks like on a spectrogram
A spectrogram shows which frequencies sound at each moment: time runs horizontally, frequency vertically, and brightness is the level. In a track exported from a project, the top of the spectrum usually fades out smoothly towards 20–22 kHz. An MP3 has a flat horizontal edge at that height: there’s almost nothing above it for the whole track.
| Source | Typical cutoff | Comment |
|---|---|---|
| MP3 128 kbps | ≈ 16–17 kHz | The most common upscale, clearly visible on the spectrum. |
| MP3 192 kbps | ≈ 18–19 kHz | The cutoff is higher, but still a flat line. |
| MP3 256–320 kbps | ≈ 19–20 kHz | Harder to tell from lossless, almost impossible by ear. |
| WAV/FLAC from the project | up to 20–22 kHz | A smooth fade; the top end changes along with the music. |
There are other signs too: “swirly”, bubbling artifacts on the cymbals, smeared drum attacks, holes in the spectrum as separate patches. But the most reliable one is a flat shelf at the top of the spectrum.
How to check your file
- Upload the file to the track analyzer. The Stream-Q analysis determines the real quality from the spectrum, not the extension: lossless, high (≈ 256–320 kbps), medium (≈ 160–192 kbps) or low (≤ 128 kbps), and adds an “upscale” mark if the stated format doesn’t match the spectrum.
- Open the file in any app with a spectrogram — the free Spek and Audacity (the “Spectrogram” track view) will do.
- Look at the top of the spectrum on a loud passage, such as the chorus. A flat horizontal edge at the same height throughout the track is a sign of a lossy codec.
Why it matters
- Re-compression. The platform or the stream will compress your track anyway. Compressing what’s already compressed adds new artifacts on top of the old ones — the sound gets muddy and “metallic”.
- Trust during the review. Many streamers turn on a submission rule that marks or holds back upscales. A mark in the queue isn’t the best start for a track.
- Release requirements. Distributors ask for WAV or FLAC, and some check the spectrum. An upscale can lead to a rejection or a re-upload.
- It’s a symptom. If the whole track is upscaled, there’s usually an MP3 beat or sample somewhere in the project. Better to find it before the track goes off for mixing.
Sample rate and bit depth: what actually matters
Upscaling isn’t only about codecs. A 96 kHz / 24-bit file made from a 44.1 kHz / 16-bit recording doesn’t gain any new information either: the spectrum above 22 kHz will be empty, and the lower bits will be filled with zeros or dither noise.
- The sample rate sets the upper limit of the recorded frequencies — half the rate. For music people will listen to, 44.1 or 48 kHz is enough.
- Bit depth sets the dynamic range and the level of quantization noise. 24 bits are useful for recording and mixing; for the final file 16 bits with dither is enough.
- Send the file at the settings the project was made in. There’s no need to resample “just in case”: every conversion is one more processing of the signal.
If there’s no source
Sometimes the project is lost and only an MP3 is left. Then it’s more honest to send the MP3 — at the highest quality you have — than to re-save it as a WAV. The analysis will show the real quality, and the streamer will understand what they’re looking at, without an “upscale” mark.
If the track matters and you plan to release it, it’s worth rebuilding it: find the stems, ask the beatmaker for a WAV, re-record the vocals. For a review on stream this isn’t necessary — the songwriting, arrangement and delivery can be judged from an MP3 too. Just mention in the title or a comment that the source is lost.
How not to send the wrong file
- Export the track straight from the project to WAV or FLAC. If the streamer limits the size, make a 320 kbps MP3 from that WAV — it’s an honest format, and the analysis will show “high quality”.
- When you buy a beat, ask for a WAV or trackouts (separate tracks). For most beatmakers the MP3 license is the cheapest and isn’t meant for release.
- Don’t use YouTube rips as samples in a track you plan to release — it’s a question not only of quality, but of copyright too.
- Keep your sources and stems. Raising the sample rate or bit depth (44.1 → 96 kHz, 16 → 24 bits) doesn’t add information — it’s the same upscaling, just on another axis.