app-i18n/vocotype
Linux offline voice input method based on FunASR Paraformer
Screenshots
View
Download
BrowseChangeLog
commit 0e1ab888bf370857663bbb1e87ffc74964daf47e
Author: Huang Rui <vowstar@gmail.com>
Date: Tue Jun 30 17:21:17 2026 +0800
treewide: set vocotype stack maintainer-needed
vocotype upstream is unmaintained, so I am stepping down and marking it and the
dependency packages it pulled in as maintainer-needed.
Signed-off-by: Huang Rui <vowstar@gmail.com>
commit 98e89cd8af82788cf27cfbc63bea66bc9ba9ec39
Author: Huang Rui <vowstar@gmail.com>
Date: Wed May 20 22:24:02 2026 +0800
app-i18n/vocotype: enable python3_14
Signed-off-by: Huang Rui <vowstar@gmail.com>
commit 57f34ba61445b96c63a5c603441f62d5947f0781
Author: Huang Rui <vowstar@gmail.com>
Date: Wed May 6 17:58:58 2026 +0800
app-i18n/vocotype: install systemd unit and XDG autostart unconditionally; fix audio capture
Per devmanual, small files like systemd service units shouldn't be gated by USE flags. Always install both the .service and the .desktop autostart, and have the .desktop's Exec defer to systemctl when available with a direct-exec fallback. This stops the GNOME session and the systemd unit from each spawning their own backend racing on /tmp/vocotype-fcitx5.sock, which produced garbled recognition. Also drop the now-redundant systemctl elog in postinst since autostart handles it.
Add vocotype-2.1.3-audio-fixes.patch covering the deeper recognition bug. PortAudio on Gentoo has no native PipeWire/PulseAudio host API, so sounddevice only sees ALSA devices. The previous fallback grabbed the first ALSA capture card (typically hw:0,0) instead of the PipeWire-injected "default" / "pulse" virtual PCM that follows the user's wpctl/pavucontrol selection. Recording the wrong jack and then linearly resampling 44.1 kHz to 16 kHz with np.interp (no anti-aliasing) produced upstream issue #5. The patch now resolves the default mic with priority "default" -> "pulse" -> sd.default.device[0] -> first input, asks the sound server for 16 kHz directly so PipeWire resamples cleanly, and falls back to librosa.resample(res_type="soxr_hq") when the server insists on its native rate. Also install scripts/setup-audio.py as vocotype-setup-audio for users who need to pin a specific microphone, and surface wpctl/pavucontrol diagnostics in postinst.
Signed-off-by: Huang Rui <vowstar@gmail.com>
commit 69c982c580e9bb7bc7ec8b2e0f776de6390a6fc3
Author: Huang Rui <vowstar@gmail.com>
Date: Tue Apr 28 21:59:20 2026 +0800
app-i18n/vocotype: add 2.1.3, drop 2.1.2
Signed-off-by: Huang Rui <vowstar@gmail.com>
commit 475c8cd8b40d5fa1e2cb70366a549325acfb4176
Author: Huang Rui <vowstar@gmail.com>
Date: Tue Mar 3 13:19:59 2026 +0800
app-i18n/vocotype: new package, add 2.1.2
Signed-off-by: Huang Rui <vowstar@gmail.com>
Author: Huang Rui <vowstar@gmail.com>
Date: Tue Jun 30 17:21:17 2026 +0800
treewide: set vocotype stack maintainer-needed
vocotype upstream is unmaintained, so I am stepping down and marking it and the
dependency packages it pulled in as maintainer-needed.
Signed-off-by: Huang Rui <vowstar@gmail.com>
commit 98e89cd8af82788cf27cfbc63bea66bc9ba9ec39
Author: Huang Rui <vowstar@gmail.com>
Date: Wed May 20 22:24:02 2026 +0800
app-i18n/vocotype: enable python3_14
Signed-off-by: Huang Rui <vowstar@gmail.com>
commit 57f34ba61445b96c63a5c603441f62d5947f0781
Author: Huang Rui <vowstar@gmail.com>
Date: Wed May 6 17:58:58 2026 +0800
app-i18n/vocotype: install systemd unit and XDG autostart unconditionally; fix audio capture
Per devmanual, small files like systemd service units shouldn't be gated by USE flags. Always install both the .service and the .desktop autostart, and have the .desktop's Exec defer to systemctl when available with a direct-exec fallback. This stops the GNOME session and the systemd unit from each spawning their own backend racing on /tmp/vocotype-fcitx5.sock, which produced garbled recognition. Also drop the now-redundant systemctl elog in postinst since autostart handles it.
Add vocotype-2.1.3-audio-fixes.patch covering the deeper recognition bug. PortAudio on Gentoo has no native PipeWire/PulseAudio host API, so sounddevice only sees ALSA devices. The previous fallback grabbed the first ALSA capture card (typically hw:0,0) instead of the PipeWire-injected "default" / "pulse" virtual PCM that follows the user's wpctl/pavucontrol selection. Recording the wrong jack and then linearly resampling 44.1 kHz to 16 kHz with np.interp (no anti-aliasing) produced upstream issue #5. The patch now resolves the default mic with priority "default" -> "pulse" -> sd.default.device[0] -> first input, asks the sound server for 16 kHz directly so PipeWire resamples cleanly, and falls back to librosa.resample(res_type="soxr_hq") when the server insists on its native rate. Also install scripts/setup-audio.py as vocotype-setup-audio for users who need to pin a specific microphone, and surface wpctl/pavucontrol diagnostics in postinst.
Signed-off-by: Huang Rui <vowstar@gmail.com>
commit 69c982c580e9bb7bc7ec8b2e0f776de6390a6fc3
Author: Huang Rui <vowstar@gmail.com>
Date: Tue Apr 28 21:59:20 2026 +0800
app-i18n/vocotype: add 2.1.3, drop 2.1.2
Signed-off-by: Huang Rui <vowstar@gmail.com>
commit 475c8cd8b40d5fa1e2cb70366a549325acfb4176
Author: Huang Rui <vowstar@gmail.com>
Date: Tue Mar 3 13:19:59 2026 +0800
app-i18n/vocotype: new package, add 2.1.2
Signed-off-by: Huang Rui <vowstar@gmail.com>