Cargo.lock

@@ -1161,6 +1161,7 @@ dependencies = [
  "common",
  "gpui-kit",
  "gpui-pre",
+ "instant",
  "itertools 0.13.0",
  "nostr-sdk",
  "person",
@@ -5430,6 +5431,7 @@ dependencies = [
  "common",
  "gpui-kit",
  "gpui-pre",
+ "instant",
  "log",
  "nip29",
  "nostr-sdk",
@@ -8374,6 +8376,7 @@ dependencies = [
  "flume 0.11.1",
  "gpui-kit",
  "gpui-pre",
+ "instant",
  "js-sys",
  "log",
  "ort",

crates/chat_ui/Cargo.toml

@@ -15,6 +15,7 @@ chat = { path = "../chat" }
 settings = { path = "../settings" }
 
 gpui.workspace = true
+instant.workspace = true
 nostr-sdk.workspace = true
 anyhow.workspace = true
 itertools.workspace = true

crates/chat_ui/src/lib.rs

@@ -96,6 +96,9 @@ pub struct ChatPanel {
     /// Dictation session for the chat input
     speech: Entity<SpeechState>,
 
+    /// When the running dictation started, for the elapsed counter
+    speech_started: Option<instant::Instant>,
+
     /// Subject input state
     subject_input: Entity<InputState>,
 
@@ -207,14 +210,25 @@ impl ChatPanel {
             // Insert the dictated text into the chat input
             cx.subscribe_in(&speech, window, move |this, _speech, event, window, cx| {
                 match event {
-                    SpeechEvent::Final(text) if !text.is_empty() => {
+                    SpeechEvent::Started => {
-                        this.input.update(cx, |input, cx| {
+                        this.speech_started = Some(instant::Instant::now());
-                            input.insert(text.clone(), window, cx);
+                        cx.notify();
-                        });
+                    }
+                    SpeechEvent::Final(text) => {
+                        this.speech_started = None;
+
+                        if !text.is_empty() {
+                            this.input
+                                .update(cx, |input, cx| input.insert(text.clone(), window, cx));
+                        }
                     }
                     SpeechEvent::Error(error) => {
+                        this.speech_started = None;
                         window.push_notification(Notification::error(error.to_string()), cx);
                     }
+                    SpeechEvent::Cancelled => {
+                        this.speech_started = None;
+                    }
                     _ => {}
                 };
             }),
@@ -238,6 +252,7 @@ impl ChatPanel {
             list_state,
             input,
             speech,
+            speech_started: None,
             subject_input,
             subject_bar,
             replies_to,
@@ -1706,7 +1721,12 @@ impl ChatPanel {
         items
     }
 
-    fn render_config_menu(&self, _window: &mut Window, cx: &Context<Self>) -> impl IntoElement {
+    fn render_config_menu(
+        &self,
+        disabled: bool,
+        _window: &mut Window,
+        cx: &Context<Self>,
+    ) -> impl IntoElement {
         let (backup, signer_kind) = self
             .room
             .read_with(cx, |this, _cx| {
@@ -1719,6 +1739,7 @@ impl ChatPanel {
             .icon(IconName::Settings2)
             .tooltip("Configuration")
             .ghost()
+            .disabled(disabled)
             .dropdown_menu(move |this, _window, _cx| {
                 let auto = matches!(signer_kind, SignerKind::Auto);
                 let encryption = matches!(signer_kind, SignerKind::Encryption);
@@ -1749,7 +1770,12 @@ impl ChatPanel {
             })
     }
 
-    fn render_emoji_menu(&self, _window: &Window, cx: &Context<Self>) -> impl IntoElement {
+    fn render_emoji_menu(
+        &self,
+        disabled: bool,
+        _window: &Window,
+        cx: &Context<Self>,
+    ) -> impl IntoElement {
         let nostr = NostrRegistry::global(cx);
         let current_user = nostr.read(cx).current_user();
 
@@ -1766,6 +1792,7 @@ impl ChatPanel {
         Button::new("emoji")
             .icon(IconName::Emoji)
             .ghost()
+            .disabled(disabled)
             .dropdown_menu_with_anchor(gpui::Anchor::BottomLeft, move |this, _window, _cx| {
                 let menu = REACTION_EMOJIS.iter().fold(this, |this, emoji| {
                     this.menu(*emoji, Box::new(Command::Insert(emoji.to_string())))
@@ -1839,7 +1866,7 @@ impl Render for ChatPanel {
 
         let is_typing = !self.input.read(cx).value().trim().is_empty();
         let pending_attachments = !self.encrypted_attachments.read(cx).is_empty();
-        let capturing = self.speech.read(cx).status().is_capturing();
+        let dictating = self.speech.read(cx).status().is_active();
 
         v_flex()
             .image_cache(retain_all(self.id.clone()))
@@ -1920,13 +1947,27 @@ impl Render for ChatPanel {
                                     .icon(IconName::Paperclip)
                                     .tooltip("Upload media")
                                     .loading(self.uploading)
-                                    .disabled(self.uploading)
+                                    .disabled(dictating || self.uploading)
                                     .ghost()
                                     .on_click(cx.listener(move |this, _ev, window, cx| {
                                         this.upload(window, cx);
                                     })),
                             )
-                            .child(
+                            .when(dictating, |this| {
+                                this.child(
+                                    h_flex()
+                                        .h_8()
+                                        .flex_1()
+                                        .gap_2()
+                                        .pl_1()
+                                        .child(SpeechWaveform::new(&self.speech).small().flex_1())
+                                        .when_some(self.speech_started, |this, started| {
+                                            this.child(speech::elapsed(started, cx))
+                                        }),
+                                )
+                            })
+                            .when(!dictating, |this| {
+                                this.child(
                                     Textarea::new(&self.input)
                                         .appearance(false)
                                         .small()
@@ -1934,28 +1975,24 @@ impl Render for ChatPanel {
                                         .py_1()
                                         .flex_1(),
                                 )
+                            })
                             .child(
                                 h_flex()
                                     .items_end()
                                     .gap_1()
                                     .pl_1()
-                                    .when(capturing, |this| {
-                                        this.child(
-                                            SpeechWaveform::new(&self.speech).xsmall().w_12(),
-                                        )
-                                    })
                                     .child(SpeechButton::new(&self.speech)),
                             )
                             .child(
                                 h_flex()
                                     .pl_1()
                                     .gap_1()
-                                    .child(self.render_emoji_menu(window, cx))
+                                    .child(self.render_emoji_menu(dictating, window, cx))
-                                    .child(self.render_config_menu(window, cx))
+                                    .child(self.render_config_menu(dictating, window, cx))
                                     .child(
                                         Button::new("send")
                                             .icon(IconName::PaperPlaneFill)
-                                            .disabled(self.uploading)
+                                            .disabled(dictating || self.uploading)
                                             .ghost()
                                             .on_click(cx.listener(move |this, _ev, window, cx| {
                                                 this.send_text_message(window, cx);

crates/nip29_ui/Cargo.toml

@@ -17,6 +17,7 @@ gpui-kit = { workspace = true, features = ["speech"] }
 ui = { path = "../ui" }
 
 gpui.workspace = true
+instant.workspace = true
 nostr-sdk.workspace = true
 anyhow.workspace = true
 log.workspace = true

crates/nip29_ui/src/lib.rs

@@ -316,6 +316,8 @@ pub struct GroupPanel {
     input: Entity<TextareaState>,
     /// Dictation session for the composer input
     speech: Entity<SpeechState>,
+    /// When the running dictation started, for the elapsed counter
+    speech_started: Option<instant::Instant>,
     /// Uploaded, non-encrypted attachments waiting to be sent.
     attachments: Vec<Url>,
     /// Whether an upload is in flight.
@@ -367,14 +369,26 @@ impl GroupPanel {
         subscriptions.push(
             cx.subscribe_in(&speech, window, |this, _, event, window, cx| {
                 match event {
-                    SpeechEvent::Final(text) if !text.is_empty() => {
+                    SpeechEvent::Started => {
+                        this.speech_started = Some(instant::Instant::now());
+                        cx.notify();
+                    }
+                    SpeechEvent::Final(text) => {
+                        this.speech_started = None;
+
+                        if !text.is_empty() {
                             this.input.update(cx, |input, cx| {
                                 input.insert(text.clone(), window, cx);
                             });
                         }
+                    }
                     SpeechEvent::Error(error) => {
+                        this.speech_started = None;
                         window.push_notification(Notification::error(error.to_string()), cx);
                     }
+                    SpeechEvent::Cancelled => {
+                        this.speech_started = None;
+                    }
                     _ => {}
                 };
             }),
@@ -416,6 +430,7 @@ impl GroupPanel {
             rendered_texts_by_id: BTreeMap::new(),
             input,
             speech,
+            speech_started: None,
             attachments: Vec::new(),
             uploading: false,
             tasks: smallvec![],
@@ -1073,6 +1088,8 @@ impl GroupPanel {
             return self.render_notice(&notice, cx);
         }
 
+        let dictating = self.speech.read(cx).status().is_active();
+
         v_flex()
             .flex_shrink_0()
             .w_full()
@@ -1091,13 +1108,26 @@ impl GroupPanel {
                             .icon(ui::IconName::Paperclip)
                             .tooltip("Upload media")
                             .loading(self.uploading)
-                            .disabled(self.uploading)
+                            .disabled(dictating || self.uploading)
                             .ghost()
                             .on_click(cx.listener(|this, _event, window, cx| {
                                 this.upload(window, cx);
                             })),
                     )
-                    .child(
+                    .when(dictating, |this| {
+                        this.child(
+                            h_flex()
+                                .h_8()
+                                .flex_1()
+                                .gap_2()
+                                .child(SpeechWaveform::new(&self.speech).small().flex_1())
+                                .when_some(self.speech_started, |this, started| {
+                                    this.child(speech::elapsed(started, cx))
+                                }),
+                        )
+                    })
+                    .when(!dictating, |this| {
+                        this.child(
                             Textarea::new(&self.input)
                                 .appearance(false)
                                 .bg(cx.theme().input)
@@ -1106,8 +1136,6 @@ impl GroupPanel {
                                 .py_1()
                                 .flex_1(),
                         )
-                    .when(self.speech.read(cx).status().is_capturing(), |this| {
-                        this.child(SpeechWaveform::new(&self.speech).xsmall().w_12())
                     })
                     .child(SpeechButton::new(&self.speech))
                     .child(
@@ -1115,7 +1143,7 @@ impl GroupPanel {
                             .icon(ui::IconName::PaperPlaneFill)
                             .tooltip("Send")
                             .ghost()
-                            .disabled(self.uploading)
+                            .disabled(dictating || self.uploading)
                             .on_click(cx.listener(|this, _event, window, cx| {
                                 this.send(window, cx);
                             })),

crates/speech/Cargo.toml

@@ -10,6 +10,7 @@ log.workspace = true
 gpui.workspace = true
 gpui-kit = { workspace = true, features = ["speech"] }
 flume.workspace = true
+instant.workspace = true
 
 [target.'cfg(not(target_arch = "wasm32"))'.dependencies]
 common = { path = "../common" }

crates/speech/src/lib.rs

@@ -5,8 +5,35 @@ mod parakeet;
 #[cfg(target_arch = "wasm32")]
 mod web;
 
+use std::time::Duration;
+
+use gpui::{
+    Animation, AnimationExt as _, AnyElement, App, IntoElement as _, ParentElement as _,
+    Styled as _, div,
+};
+use gpui_kit::component::ActiveTheme as _;
 use gpui_kit::component::speech::SpeechRecognizer;
 
+/// A ticking `M:SS` counter for the dictation in progress.
+pub fn elapsed(started: instant::Instant, cx: &App) -> AnyElement {
+    div()
+        .text_xs()
+        .text_color(cx.theme().muted_foreground)
+        .with_animation(
+            "dictation-elapsed",
+            Animation::new(Duration::from_secs(1))
+                .repeat()
+                .with_max_fps(1.),
+            move |counter, _| counter.child(format_elapsed(started)),
+        )
+        .into_any_element()
+}
+
+fn format_elapsed(started: instant::Instant) -> String {
+    let seconds = started.elapsed().as_secs();
+    format!("{}:{:02}", seconds / 60, seconds % 60)
+}
+
 /// The recognizer backing the composer's microphone button.
 #[cfg(not(target_arch = "wasm32"))]
 pub fn recognizer() -> impl SpeechRecognizer {