The size is arithmetic, not a guess
Output is constant-bitrate MP3, so the size is simply duration × bitrate ÷ 8. A four-minute track at 128 kbps is about 3.8 MB regardless of what it contains. That is why the estimate appears before you encode, and why it is accurate rather than approximate.
It also means the warning is meaningful: if your source is already a 96 kbps MP3, re-encoding at 192 makes a bigger file that sounds slightly worse. The page says so rather than letting you find out afterwards.
Choosing a bitrate honestly
320 is transparent for essentially everyone on essentially everything. 192 is high quality and half the size. 128 is the long-standing default and fine on phone speakers and earbuds. 96 starts to show on cymbals and applause. 64 is for speech only, where it is genuinely acceptable.
For a podcast, 64 kbps mono is a legitimate choice that many large shows use — it is a quarter the size of 128 stereo and no listener notices.
Every re-encode costs something
MP3 is lossy, so decoding and re-encoding always loses a little more, even at a higher bitrate. If you still have the original WAV or FLAC, compress from that instead of from an MP3. If you only have the MP3, one more generation is usually fine — it is repeated round trips that audibly degrade.
Frequently asked questions
Why is the estimate slightly different from the real file?
The encoder adds a small header and pads the final frame, which is a few kilobytes. On anything longer than a few seconds the difference is under half a percent.
Is my audio uploaded?
No. The re-encoding happens on your own machine — the file never leaves your device.
Something wrong with this tool, or an idea for it? Tell us