Flatpak build

jomason · Dec 1, 2022 · 563e7be · 563e7be
1 parent bb6745b
commit 563e7be
Show file tree

Hide file tree

Showing 14 changed files with 284 additions and 134 deletions.
diff --git a/.gitignore b/.gitignore
@@ -0,0 +1,4 @@
+build
+build/*
+_build
+_build/*
diff --git a/.gitmodules b/.gitmodules
@@ -0,0 +1,3 @@
+[submodule "subprojects/april-asr"]
+	path = subprojects/april-asr
+	url = git@github.com:abb128/april-asr.git
diff --git a/data/net.sapples.LiveCaptions.gschema.xml b/data/net.sapples.LiveCaptions.gschema.xml
@@ -1,5 +1,39 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <schemalist gettext-domain="livecaptions">
-	<schema id="net.sapples.LiveCaptions" path="/net/sapples/LiveCaptions/">
-	</schema>
+    <schema id='net.sapples.LiveCaptions' path='/net/sapples/LiveCaptions/'>
+        <key name="font-name" type="s">
+            <default>'Sans Regular'</default>
+            <summary>The font of the captions</summary>
+        </key>
+
+        <key name="font-size" type="i">
+            <range min="1" max="144" />
+            <default>12</default>
+            <summary>The font size of the captions</summary>
+        </key>
+
+        <key name="text-uppercase" type="b">
+            <default>false</default>
+            <summary>Display letters in uppercase instead of lowercase</summary>
+        </key>
+
+        <key name="fade-text" type="b">
+            <default>true</default>
+            <summary>Fade text based on confidence</summary>
+        </key>
+
+        <key name="filter-profanity" type="b">
+            <default>true</default>
+            <summary>Filter swear words and slurs</summary>
+        </key>
+
+        <key name="microphone" type="b">
+            <default>false</default>
+            <summary>Caption microphone input instead of desktop audio</summary>
+        </key>
+
+        <key name="benchmark" type="d">
+            <default>-1.0</default>
+        </key>
+    </schema>
 </schemalist>
diff --git a/meson.build b/meson.build
@@ -7,6 +7,11 @@ project('livecaptions', 'c',
                    ],
 )
 
+cmake = import('cmake')
+april = cmake.subproject('april-asr')
+
+april_lib = april.dependency('aprilasr_static')
+
 i18n = import('i18n')
 
 gnome = import('gnome')

diff --git a/net.sapples.LiveCaptions.json b/net.sapples.LiveCaptions.json
@@ -12,7 +12,7 @@
         "--socket=pulseaudio",
         "--socket=wayland",
         "--filesystem=xdg-run/pipewire-0:ro",
-        "--env=LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/app/lib/pipewire-0.3/jack"
+        "--env=LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/app/lib/pipewire-0.3/jack:/app/lib"
     ],
     "cleanup" : [
         "/include",
@@ -22,10 +22,32 @@
         "/share/gtk-doc",
         "/share/man",
         "/share/pkgconfig",
+        "/onnxruntime_ROOT",
         "*.la",
         "*.a"
     ],
     "modules" : [
+        {
+            "name" : "onnxruntime",
+            "buildsystem" : "simple",
+            "sources" : [
+                {
+                    "type" : "file",
+                    "dest-filename" : "microsoft.ml.onnxruntime.1.13.1.nupkg",
+                    "url" : "https://globalcdn.nuget.org/packages/microsoft.ml.onnxruntime.1.13.1.nupkg",
+                    "sha256" : "590e655b036e6e663ad1274bf8650ae671859005cd61eff432dfd3dba66b5904"
+            	}
+            ],
+            "build-commands" : [
+                "mkdir ${FLATPAK_DEST}/onnxruntime_ROOT",
+                "mv microsoft.ml.onnxruntime.1.13.1.nupkg ${FLATPAK_DEST}/onnxruntime_ROOT/Microsoft.ML.OnnxRuntime.1.13.1.nupkg",
+                "cd ${FLATPAK_DEST}/onnxruntime_ROOT && unzip Microsoft.ML.OnnxRuntime.1.13.1.nupkg",
+
+                "install -Dm755 ${FLATPAK_DEST}/onnxruntime_ROOT/runtimes/linux-x64/native/libonnxruntime.so ${FLATPAK_DEST}/lib/libonnxruntime.so.1.13.1",
+                "ln -s ${FLATPAK_DEST}/lib/libonnxruntime.so.1.13.1 ${FLATPAK_DEST}/lib/libonnxruntime.so"
+            ]
+        },
+
         {
             "name" : "livecaptions",
             "builddir" : true,
@@ -36,6 +58,23 @@
                     "url" : "file:///home/alex/Projects/LiveCaptions"
                 }
             ]
+        },
+
+        {
+            "name": "april-model-v0",
+            "buildsystem" : "simple",
+            "sources" : [
+                {
+                    "type" : "file",
+                    "dest-filename" : "aprilv0_en-us.april",
+                    "url" : "https://april.sapples.net/aprilv0_en-us.april",
+                    "sha256" : "f5a30f888e797769b1b7de143c05d2c92a1702c263b5b71d62fd2028cede5627"
+                }
+            ],
+            "build-commands" : [
+                "mkdir -p ${XDG_DATA_HOME}/LiveCaptions/models",
+                "install -Dm644 aprilv0_en-us.april ${XDG_DATA_HOME}/LiveCaptions/models"
+            ]
         }
     ]
 }
diff --git a/src/asrproc.c b/src/asrproc.c
@@ -15,6 +15,8 @@
 #include "line-gen.h"
 
 struct asr_thread_i {
+    size_t silence_counter;
+
     FILE *fd;
 
     GThread * thread_id;
@@ -66,6 +68,21 @@ void april_result_handler(void* userdata, AprilResultType result, size_t count,
 }
 
 void asr_thread_enqueue_audio(asr_thread thread, short *data, size_t num_shorts) {
+    bool found_nonzero = false;
+    for(int i=0; i<num_shorts; i++){
+        if((data[i] > 8) || (data[i] < -8)){
+            found_nonzero = true;
+            break;
+        }
+    }
+
+    thread->silence_counter = found_nonzero ? 0 : (thread->silence_counter + num_shorts);
+
+    if(thread->silence_counter >= 24000){
+        thread->silence_counter = 24000;
+        return aas_flush(thread->session);
+    }
+
     fwrite(data, num_shorts, 2, thread->fd);
     aas_feed_pcm16(thread->session, data, num_shorts); // TODO?
 }

diff --git a/src/livecaptions-settings.c b/src/livecaptions-settings.c
@@ -16,6 +16,8 @@
  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
+#include <glib/gi18n.h>
+
 #include "livecaptions-config.h"
 #include "livecaptions-settings.h"
 #include "livecaptions-application.h"
@@ -48,8 +50,81 @@ toast_show_cb (AdwPreferencesWindow *window)
 }
 
 
+static void issue_tracker_cb(LiveCaptionsSettings *self){
+    system("xdg-open https://github.com/abb128/LiveCaptions/issues");
+}
+
+static void git_cb(LiveCaptionsSettings *self){
+    system("xdg-open https://github.com/abb128/LiveCaptions");
+}
+
+static void about_cb(LiveCaptionsSettings *self) {
+    GtkRoot *root = gtk_widget_get_root (GTK_WIDGET (self));
+    GtkWidget *about;
+
+    const char *developers[] = {
+        "abb128",
+        NULL
+    };
+
+    const char *special_thanks[] = {
+        "Fangjun Kuang (@csukuangfj) and the k2-fsa/icefall contributors",
+        NULL
+    };
+
+    //const char *release_notes = "\
+    //    <p>\
+    //    This release adds the following features:\
+    //    </p>\
+    //    <ul>\
+    //    <li>Added a way to export fonts.</li>\
+    //    <li>Better support for <code>monospace</code> fonts.</li>\
+    //    <li>Added a way to preview <em>italic</em> text.</li>\
+    //    <li>Bug fixes and performance improvements.</li>\
+    //    <li>Translation updates.</li>\
+    //    </ul>\
+    //";
+
+    about =
+        g_object_new (ADW_TYPE_ABOUT_WINDOW,
+                    "transient-for", root,
+                    "application-icon", "net.sapples.LiveCaptions",
+                    "application-name", _("Live Captions"),
+                    "developer-name", _("abb128"),
+                    "version", "0.0.1",
+                    //"release-notes-version", "1.2.0",
+                    //"release-notes", release_notes,
+                    "comments", _("Live Captions is an application for the Linux Desktop that captions desktop audio."),
+                    "website", "https://github.com/abb128/LiveCaptions",
+                    "issue-url", "https://github.com/abb128/LiveCaptions/issues",
+                    //"support-url", "https://example.org",
+                    "copyright", "© 2022 abb128",
+                    "license-type", GTK_LICENSE_GPL_3_0,
+                    "developers", developers,
+                    //"artists", artists,
+                    "translator-credits", _("translator-credits"),
+                    NULL);
+
+    //adw_about_window_add_link (ADW_ABOUT_WINDOW (about),
+    //                            _("_Documentation"),
+    //                            "https://gnome.pages.gitlab.gnome.org/libadwaita/doc/main/class.AboutWindow.html");
+
+    adw_about_window_add_legal_section (ADW_ABOUT_WINDOW (about),
+                                        _("Model"),
+                                        NULL,
+                                        GTK_LICENSE_CUSTOM,
+                                        "The ASR model was originally trained by Fangjun Kuang (@csukuangfj), and has been finetuned on extra data.");
+
+    adw_about_window_add_acknowledgement_section (ADW_ABOUT_WINDOW (about),
+                                                    _("Special thanks to"),
+                                                    special_thanks);
+
+    gtk_window_present (GTK_WINDOW (about));
+}
+
+
 static void change_font(LiveCaptionsSettings *self) {
-    
+
 }
 
 static void livecaptions_settings_class_init(LiveCaptionsSettingsClass *klass) {
@@ -59,8 +134,17 @@ static void livecaptions_settings_class_init(LiveCaptionsSettingsClass *klass) {
     gtk_widget_class_bind_template_child (widget_class, LiveCaptionsSettings, subpage1);
     gtk_widget_class_bind_template_child (widget_class, LiveCaptionsSettings, subpage2);
 
+    gtk_widget_class_bind_template_child (widget_class, LiveCaptionsSettings, font_button);
+    gtk_widget_class_bind_template_child (widget_class, LiveCaptionsSettings, text_upper_switch);
+    gtk_widget_class_bind_template_child (widget_class, LiveCaptionsSettings, fade_text_switch);
+    gtk_widget_class_bind_template_child (widget_class, LiveCaptionsSettings, filter_profanity_switch);
+
+
     gtk_widget_class_bind_template_callback (widget_class, change_font);
     gtk_widget_class_bind_template_callback (widget_class, return_to_preferences_cb);
+    gtk_widget_class_bind_template_callback (widget_class, issue_tracker_cb);
+    gtk_widget_class_bind_template_callback (widget_class, git_cb);
+    gtk_widget_class_bind_template_callback (widget_class, about_cb);
     gtk_widget_class_bind_template_callback (widget_class, subpage1_activated_cb);
     gtk_widget_class_bind_template_callback (widget_class, subpage2_activated_cb);
 
@@ -69,4 +153,11 @@ static void livecaptions_settings_class_init(LiveCaptionsSettingsClass *klass) {
 
 static void livecaptions_settings_init(LiveCaptionsSettings *self) {
     gtk_widget_init_template(GTK_WIDGET(self));
+
+    self->settings = g_settings_new("net.sapples.LiveCaptions");
+
+    g_settings_bind(self->settings, "text-uppercase", self->text_upper_switch, "state", G_SETTINGS_BIND_DEFAULT);
+    g_settings_bind(self->settings, "fade-text", self->fade_text_switch, "state", G_SETTINGS_BIND_DEFAULT);
+    g_settings_bind(self->settings, "filter-profanity", self->filter_profanity_switch, "state", G_SETTINGS_BIND_DEFAULT);
+
 }
diff --git a/src/livecaptions-settings.h b/src/livecaptions-settings.h
@@ -23,6 +23,13 @@
 struct _LiveCaptionsSettings {
     AdwPreferencesWindow  parent_instance;
 
+    GSettings *settings;
+
+    GtkFontButton *font_button;
+    GtkSwitch *text_upper_switch;
+    GtkSwitch *fade_text_switch;
+    GtkSwitch *filter_profanity_switch;
+
     GtkWidget *subpage1;
     GtkWidget *subpage2;