mp3rgui now speaks five languages: English, Brazilian Portuguese, Italian, Japanese, and Turkish. The translation system behind them was built by @cfgnunes. On the command line, the analysis and the gains that -r, -a, and -g apply are unchanged. What changed is how the stored mp3gain values are written (the global_gain ranges, the album range, and the undo record), each time toward what mp3gain 1.6.2 writes, and -a -r now applies track gain, as in mp3gain.
mp3rgui speaks five languages (#422, [#439])
@cfgnunes designed and built mp3rgui's translation system (#439):
- Each language is one gettext
.pofile, compiled into the app at build time. A translation ships in every package (installer, DMG, winget, Homebrew, apt, AUR) with nothing extra to install. - The English text is the key, so a missing or outdated entry shows English, string by string. A release never waits for a translation.
- A new Language menu offers System default, English, and every translation under its own name, and remembers the choice.
- A CI check rejects unfinished (
fuzzy) entries and placeholders that do not match the English, so a translation can be merged on CI alone.
| Language | File | Translated by |
|---|---|---|
| Português (Brasil) | pt_BR.po |
@cfgnunes (#439) |
| Italiano | it.po |
@Privato72, from their own Italian build of mp3rgui in [#406] (#443) |
| 日本語 | ja.po |
@M-Igashi, with Claude Code (Opus 5.5) as co-author (#445) |
| Türkçe | tr.po |
@0anes0 (#448, [#457]) |
The command line stays English: scripts and mp3gain front-ends read its output.
Want mp3rgui in your language? It is one text file and needs no programming: see [#440]. Portuguese (Portugal) is already offered. Arabic, Hebrew, Persian, Thai, and the Indic scripts have to wait for an upgrade of the GUI toolkit, which cannot shape them yet.
The font for Chinese, Japanese, and Korean file names now follows the system language on macOS and Windows too, where it used to read only the POSIX locale variables. When you pick a UI language that is not one of the three, it keeps following the system language (#452).
Behaviour changes
| What | 3.13.0 | 3.14.0 |
|---|---|---|
-a -r |
Album gain | Track gain: the later of -a and -r wins, as in mp3gain |
-l -w |
Saturated, undo recorded with N |
Wraps, undo recorded with W, and -u honours it |
MP3GAIN_MINMAX / MP3GAIN_ALBUM_MINMAX below 100 |
94,168 |
094,168, as mp3gain writes and reads it |
-l |
Rewrote or added MP3GAIN_MINMAX |
Leaves it as it is |
-g on a file without a stored range |
Added MP3GAIN_MINMAX |
Adds none, and shifts a stored one |
-r / -a on a file that needs no change |
No MP3GAIN_MINMAX |
Records it |
-a -p |
Lost the modification time | Keeps it |
-s i on a file whose mp3gain values are in APEv2 |
Started a second undo record | Moves the values into ID3v2, as mp3gain does |
-a -s i |
No album range | Album range in ID3v2 |
Undo after mixing tag layouts (#449)
An apply under -s i read the undo record from ID3v2 only. On a file whose record was in APEv2 (from mp3gain, -s a, or the default layout), it started a second record in ID3v2 from zero, and -u then restored neither the original nor the previous state. With the default layout the file could end up louder than the original. Now -s i moves the mp3gain values out of APEv2 into ID3v2 and continues from them, as mp3gain's -s i does, keeping every other APEv2 item. mp3rgui's "All in ID3v2" setting follows.
Files already in that state carry two undo records: mp3rgain -s c file.mp3 shows the APEv2 one and mp3rgain -s c -s i file.mp3 the ID3v2 one. Running mp3rgain -u twice with 3.14.0 restores the original audio exactly.
Stored global_gain ranges, as mp3gain writes them (#432, [#433], [#437], [#450], [#451])
- Three-digit padding (#432). mp3gain stores the ranges as
%03d,%03dand reads them at fixed offsets. A minimum below 100, written unpadded as94,168, made mp3gain read the maximum as 68. Ranges written by earlier versions still read correctly in mp3rgain, and mp3gain reads them correctly once an apply has written them again. -lleaves the range alone (#433). mp3gain updates the stored values only when both channels move by the same amount. With-s athe file-lwrites is now byte-identical to mp3gain's.-gshifts a stored range and adds none (#451). Under-w, a range the shift would take outside 0..255 is dropped, as in mp3gain. Inside those bounds mp3rgain shifts it, where mp3gain 1.6.2 leaves it describing the audio before the change. On a file without a stored range,-gnow writes the same bytes as mp3gain.- A run that needs no change records the range (#437).
-rand-aon a file already on target storeMP3GAIN_MINMAX, so a later mp3gain run can reuse it instead of re-analyzing. Still no undo value, since nothing changed. -a -s iwrites the album range (#450) as an ID3v2 frame, as mp3gain does.-s cshows it, and-s dand-uremove it.
-l -w wraps (#434, [#438])
mp3gain applies -w to a one-channel change too: a left global_gain of 250 plus 10 steps becomes 4, and the undo record says W. mp3rgain saturated at 255 and recorded N, and the undo of unequal left and right values always saturated, so a -l -w change made by mp3gain was not reversed. Both are fixed by @dannyward630: -l 0 10 -w now writes the same audio as mp3gain, and -u reverses mp3gain's W records exactly. Without -w nothing changes. In mp3rgui, Apply Channel Gain now follows the Wrap mode option, which it used to ignore.
-a -r applies track gain (#435)
In mp3gain, -a and -r each clear the other, so the later one wins. mp3rgain applied album gain whenever both were given. -a -r now applies each file's track gain and -r -a the album gain, byte-identical to mp3gain. mp3gain does not accept combined flags such as -ar; mp3rgain reads them left to right. -a -e still applies track gain (#378).
-a -p keeps the modification time (#436)
The album range is written in a second pass once the whole album has been applied, and that write left each file with the current time. The time is now restored after it, in the CLI and in mp3rgui's Apply Album Gain with Preserve mtime.
For library users
- New:
write_album_minmax_with_layout(files, layout)writes the album range in the container the layout uses.write_album_minmaxkeeps its signature and writes APEv2. GainOptions::wrapnow applies to channel gain too (#438).ape::format_minmaxpads both numbers to three digits.
Internal
- CI runs mp3rgui's tests, including the catalogue checks, on all three platforms (#439).
- The apt installs in CI time out after 10 minutes and skip recommended packages, after a slow Ubuntu mirror held them for up to 37 minutes (#453).
Contributors
Thank you to everyone who made this release:
- @cfgnunes built the translation system and translated mp3rgui into Brazilian Portuguese (#439). They also reported [#432], [#433], [#434], [#435], [#436], and [#437], each with mp3gain's source already read, which is why every one of those fixes could match mp3gain exactly.
- @Privato72 translated mp3rgui into Italian, starting from their own Italian build in [#406] (#443).
- @0anes0 (Ahmet Enes Kaymak) translated mp3rgui into Turkish (#448, [#457]).
- @dannyward630 fixed wrap mode for channel gain (#438).
- @M-Igashi translated mp3rgui into Japanese, with Claude Code (Opus 5.5) as co-author (#445), and made the other fixes.
Full changelog: https://github.com/M-Igashi/mp3rgain/compare/v3.13.0...v3.14.0