feat: add text to speech (#69)

Reviewed-on: #69

Cargo.lock

@@ -2155,6 +2155,27 @@ dependencies = [
  "wio",
 ]
 
+[[package]]
+name = "dyn-clonable"
+version = "0.9.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a36efbb9bfd58e1723780aa04b61aba95ace6a05d9ffabfdb0b43672552f0805"
+dependencies = [
+ "dyn-clonable-impl",
+ "dyn-clone",
+]
+
+[[package]]
+name = "dyn-clonable-impl"
+version = "0.9.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7e8671d54058979a37a26f3511fbf8d198ba1aa35ffb202c42587d918d77213a"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.119",
+]
+
 [[package]]
 name = "dyn-clone"
 version = "1.0.20"
@@ -5939,6 +5960,8 @@ version = "0.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "13a380031deed8e99db00065c45937da434ca987c034e13b87e4441f9e4090be"
 dependencies = [
+ "bitflags 2.13.2",
+ "block2 0.6.2",
  "objc2 0.6.5",
  "objc2-foundation 0.3.2",
 ]
@@ -6436,6 +6459,15 @@ dependencies = [
  "ureq",
 ]
 
+[[package]]
+name = "oxilangtag"
+version = "0.1.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5d3b4eb570abd4a1dcb062c31fd37b832264d9dc7292c3e69acfe926c87b063f"
+dependencies = [
+ "serde",
+]
+
 [[package]]
 name = "page_size"
 version = "0.6.0"
@@ -8372,6 +8404,7 @@ dependencies = [
  "serde",
  "serde_json",
  "sha2 0.10.9",
+ "tts",
  "wasm-bindgen",
  "web-sys",
 ]
@@ -8443,6 +8476,21 @@ dependencies = [
  "web-sys",
 ]
 
+[[package]]
+name = "ssip"
+version = "0.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b81f5ada7d4745722e62f8be4aca56abcf0e0e46e409612ac94a4a3a2b471d39"
+
+[[package]]
+name = "ssip-client-async"
+version = "0.19.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2cfbcb0ed774006ab9f14d4595fc02668cd7f5e662a87ff82264e6506238f47f"
+dependencies = [
+ "ssip",
+]
+
 [[package]]
 name = "stable_deref_trait"
 version = "1.2.1"
@@ -9289,6 +9337,33 @@ dependencies = [
  "core_maths",
 ]
 
+[[package]]
+name = "tts"
+version = "0.27.0"
+source = "git+https://github.com/ndarilek/tts-rs#8fbcb720cb86c5166a9ed46272e3777f742da18e"
+dependencies = [
+ "block2 0.6.2",
+ "dyn-clonable",
+ "hound",
+ "jni 0.22.4",
+ "libloading",
+ "ndk-context",
+ "objc2 0.6.5",
+ "objc2-avf-audio",
+ "objc2-foundation 0.3.2",
+ "oxilangtag",
+ "parking_lot",
+ "ssip",
+ "ssip-client-async",
+ "tempfile",
+ "thiserror 2.0.21",
+ "tracing",
+ "wasm-bindgen",
+ "web-sys",
+ "windows 0.62.2",
+ "zbus",
+]
+
 [[package]]
 name = "tungstenite"
 version = "0.28.0"

Cargo.toml

@@ -50,6 +50,7 @@ rand = { version = "0.10", default-features = false, features = [ "std", "sys_rn
 
 # Others
 anyhow = "1.0.44"
+tts = { git = "https://github.com/ndarilek/tts-rs" }
 chrono = { version = "0.4.38", features = ["wasmbind"] }
 futures = "0.3"
 itertools = "0.13.0"

assets/icons/speech.svg

@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none">
+<path d="M10.75 14.75V17.25" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/><path d="M14.25 7.75V20.25" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/><path d="M17.75 10.75V17.25" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/><path d="M21.25 12.75V15.25" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/><path d="M2.75 5.25V3.75H7M11.25 5.25V3.75H7M7 3.75V12.25M7 12.25H5.25M7 12.25H8.75" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/>
+</svg>

crates/chat_ui/src/lib.rs

@@ -30,13 +30,14 @@ use person::{Directory, Person};
 use regex::Regex;
 use settings::{AppSettings, SignerKind};
 use smallvec::{SmallVec, smallvec};
+use speech::speaker::{END_GRACE, Speaker};
 use state::{
     FileAttachment, LocalFile, NostrRegistry, PickedFile, download_and_decrypt_file, pick_file,
 };
 use ui::{
     Avatar, COMPACT_REACTION_EMOJIS, IconName, MessageActions, MessageMedia, MessageRow,
-    REACTION_EMOJIS, ReactionBar, RenderedText, ReplyPreview, WelcomeMessage, group_reactions,
+    REACTION_EMOJIS, ReactionBar, RenderedText, ReplyPreview, RevealText, WelcomeMessage,
-    opens_run, rendered_text,
+    group_reactions, opens_run, rendered_text,
 };
 
 use crate::file::*;
@@ -45,8 +46,7 @@ use crate::file::*;
 const PRIVATE_NOTICE: &str =
     "This conversation is private. Only members can see each other's messages.";
 
-/// Regex matching strings that consist entirely of emoji characters,
+/// Regex matching strings that consist entirely of emoji characters.
-/// zero-width joiners, variation selectors, and keycap combiners.
 static EMOJI_RE: LazyLock<Regex> =
     LazyLock::new(|| Regex::new(r"^[\p{Emoji}\u{200D}\u{FE0F}\u{20E3}]+$").unwrap());
 
@@ -65,65 +65,49 @@ pub fn init(room: WeakEntity<Room>, window: &mut Window, cx: &mut App) -> Entity
 pub struct ChatPanel {
     id: SharedString,
     focus_handle: FocusHandle,
-
     /// Chat Room
     room: WeakEntity<Room>,
-
     /// Person directory shared with the room
     directory: Entity<Directory>,
-
     /// Message list state
     list_state: ListState,
-
     /// All messages (sorted by created_at)
     messages: Vec<Message>,
-
     /// O(1) message lookup by EventId
     message_index: HashMap<EventId, usize>,
-
     /// All reactions
     reactions: BTreeMap<EventId, Vec<(SharedString, Option<SharedUri>, PublicKey)>>,
-
     /// Mapping message ids to their rendered texts
     rendered_texts_by_id: BTreeMap<EventId, RenderedText>,
-
     /// Mapping message (rumor event) ids to their reports
     reports_by_id: Arc<RwLock<BTreeMap<EventId, Vec<SendReport>>>>,
-
     /// Chat input state
     input: Entity<TextareaState>,
-
     /// Dictation session for the chat input
     speech: Entity<SpeechState>,
-
     /// When the running dictation started, for the elapsed counter
     speech_started: Option<instant::Instant>,
-
+    /// Reads messages aloud
+    speaker: Speaker,
+    /// The message being revealed in step with speech, with its estimated duration
+    speech_reveal: Option<(EventId, Duration)>,
+    /// Fallback reveal end for backends that cannot report utterance end
+    reveal_timeout: Option<Task<Result<(), Error>>>,
     /// Subject input state
     subject_input: Entity<InputState>,
-
     /// Subject bar visibility
     subject_bar: Entity<bool>,
-
     /// Replies to
     replies_to: Entity<HashSet<EventId>>,
-
     /// Media Attachment
     attachments: Entity<Vec<Url>>,
-
     /// Uploaded, encrypted file attachments which are not sent yet
     encrypted_attachments: Entity<Vec<PendingFile>>,
-
     /// Decrypted attachments of file messages, by message id
     decrypted_files: HashMap<EventId, DecryptedFile>,
-
     /// Upload state
     uploading: bool,
-
-    /// Async operations
     tasks: Vec<Task<Result<(), Error>>>,
-
-    /// Event subscriptions
     subscriptions: SmallVec<[Subscription; 4]>,
 }
 
@@ -241,6 +225,20 @@ impl ChatPanel {
             this.get_messages(window, cx);
         });
 
+        // End the speech reveal when the utterance finishes ahead of its estimate
+        let speaker = Speaker::new();
+        let end_events = speaker.end_events();
+        let reveal_end_task = cx.spawn(async move |this, cx| {
+            while end_events.recv_async().await.is_ok() {
+                this.update(cx, |this, cx| {
+                    this.speech_reveal = None;
+                    cx.notify();
+                })?;
+            }
+
+            Ok(())
+        });
+
         Self {
             focus_handle: cx.focus_handle(),
             id,
@@ -253,6 +251,9 @@ impl ChatPanel {
             input,
             speech,
             speech_started: None,
+            speaker,
+            speech_reveal: None,
+            reveal_timeout: None,
             subject_input,
             subject_bar,
             replies_to,
@@ -263,7 +264,7 @@ impl ChatPanel {
             reports_by_id,
             uploading: false,
             subscriptions,
-            tasks: vec![],
+            tasks: vec![reveal_end_task],
         }
     }
 
@@ -1192,6 +1193,15 @@ impl ChatPanel {
             nostr.read(cx).current_user(),
         ));
 
+        let body = match self
+            .speech_reveal
+            .as_ref()
+            .filter(|(reveal_id, _)| *reveal_id == id)
+        {
+            Some((_, duration)) => RevealText::new(message.preview(), *duration).into_any_element(),
+            None => rendered_text,
+        };
+
         MessageRow::new(ix)
             .show_author(show_author)
             .hide_avatar(hide_avatar)
@@ -1216,7 +1226,7 @@ impl ChatPanel {
             .when(has_replies, |this| {
                 this.children(self.render_message_replies(replies, cx))
             })
-            .when(message.file.is_none(), |this| this.child(rendered_text))
+            .when(message.file.is_none(), |this| this.child(body))
             .child(MessageMedia::new(&message.media))
             .when_some(message.file.as_ref(), |this, file| {
                 this.child(self.render_message_file(&id, file, cx))
@@ -1464,6 +1474,43 @@ impl ChatPanel {
             actions
         };
 
+        let speaking = self.speech_reveal.is_some();
+        let speech_id = *id;
+
+        let actions = actions.speech(
+            speaking,
+            cx.listener(move |this, _event, window, cx| {
+                let text = this.message(&speech_id).map(|message| message.preview());
+
+                if let Some(text) = text {
+                    match this.speaker.speak(&text) {
+                        Ok(duration) => {
+                            this.speech_reveal = Some((speech_id, duration));
+
+                            // Re-enable speaking when no end event ever arrives
+                            let session = (speech_id, duration);
+                            let timer = cx.background_executor().timer(duration + END_GRACE);
+                            this.reveal_timeout = Some(cx.spawn(async move |this, cx| {
+                                timer.await;
+                                this.update(cx, |this, cx| {
+                                    if this.speech_reveal.as_ref() == Some(&session) {
+                                        this.speech_reveal = None;
+                                        cx.notify();
+                                    }
+                                })?;
+
+                                Ok(())
+                            }));
+                            cx.notify();
+                        }
+                        Err(error) => {
+                            window.push_notification(Notification::error(error.to_string()), cx)
+                        }
+                    }
+                }
+            }),
+        );
+
         actions.dropdown(move |menu, _window, _cx| {
             menu.menu("Copy author", Box::new(Command::Copy(public_key)))
                 .menu("Seen on", Box::new(Command::Trace(dropdown_id)))

crates/nip29_ui/src/lib.rs

@@ -26,8 +26,9 @@ use person::Directory;
 use serde::Deserialize;
 use settings::AppSettings;
 use smallvec::{SmallVec, smallvec};
+use speech::speaker::Speaker;
 use state::{NostrRegistry, pick_file, upload_file};
-use ui::{Avatar, RenderedText, WelcomeMessage, rendered_text};
+use ui::{Avatar, RenderedText, RevealText, WelcomeMessage, rendered_text};
 use util::{attachment_name, display_name, opens_run, pin_label, report};
 
 mod details;
@@ -320,6 +321,12 @@ pub struct GroupPanel {
     speech: Entity<SpeechState>,
     /// When the running dictation started, for the elapsed counter
     speech_started: Option<instant::Instant>,
+    /// Reads messages aloud
+    speaker: Speaker,
+    /// The message being revealed in step with speech, with its estimated duration
+    speech_reveal: Option<(EventId, Duration)>,
+    /// Fallback reveal end for backends that cannot report utterance end
+    reveal_timeout: Option<Task<Result<()>>>,
     /// Uploaded, non-encrypted attachments waiting to be sent.
     attachments: Vec<Url>,
     /// Whether an upload is in flight.
@@ -419,9 +426,24 @@ impl GroupPanel {
                     }
                 },
             ));
+            this.track_participants(cx);
         });
 
-        let mut this = Self {
+        // End the speech reveal when the utterance finishes ahead of its estimate
+        let speaker = Speaker::new();
+        let end_events = speaker.end_events();
+        let reveal_end_task = cx.spawn(async move |this, cx| {
+            while end_events.recv_async().await.is_ok() {
+                this.update(cx, |this, cx| {
+                    this.speech_reveal = None;
+                    cx.notify();
+                })?;
+            }
+
+            Ok(())
+        });
+
+        Self {
             focus_handle: cx.focus_handle(),
             group: group.downgrade(),
             directory: Directory::new(cx),
@@ -433,19 +455,17 @@ impl GroupPanel {
             input,
             speech,
             speech_started: None,
+            speaker,
+            speech_reveal: None,
+            reveal_timeout: None,
             attachments: Vec::new(),
             uploading: false,
-            tasks: smallvec![],
+            tasks: smallvec![reveal_end_task],
             _subscriptions: subscriptions,
-        };
+        }
-
-        this.track_participants(cx);
-
-        this
     }
 
-    /// Track the keys rendered by this panel: members, message authors, and
+    /// Track the keys rendered by this panel.
-    /// activity subjects
     fn track_participants(&mut self, cx: &mut Context<Self>) {
         let Some(group) = self.group.upgrade() else {
             return;
@@ -1023,7 +1043,13 @@ impl GroupPanel {
         let nostr = NostrRegistry::global(cx);
         let directory = self.directory.clone();
 
-        let content = self
+        let content = match self
+            .speech_reveal
+            .as_ref()
+            .filter(|(reveal_id, _)| *reveal_id == message.id)
+        {
+            Some((_, duration)) => RevealText::new(message.preview(), *duration).into_any_element(),
+            None => self
                 .rendered_texts_by_id
                 .entry(message.id)
                 .or_insert_with(|| {
@@ -1038,7 +1064,8 @@ impl GroupPanel {
                         cx,
                     )
                 })
-            .element(ix.into(), window, cx);
+                .element(ix.into(), window, cx),
+        };
 
         let admin = nostr
             .read(cx)
@@ -1059,6 +1086,7 @@ impl GroupPanel {
                 panel: cx.weak_entity(),
                 admin,
                 pinned,
+                speaking: self.speech_reveal.is_some(),
                 persons: self.directory.clone(),
             },
             cx,

crates/nip29_ui/src/message.rs

@@ -8,15 +8,17 @@ use gpui::{
 use gpui_kit::assets::IconName;
 use gpui_kit::base::StyledExt as _;
 use gpui_kit::component::menu::{ContextMenuExt as _, PopupMenuItem};
-use gpui_kit::component::{ActiveTheme, Icon, Sizable, h_flex};
+use gpui_kit::component::notification::Notification;
+use gpui_kit::component::{ActiveTheme, Icon, Sizable, WindowExt, h_flex};
 use nip29::{Activity, Group, Reaction};
 use nostr_sdk::prelude::{EventId, Kind};
 use person::Directory;
 use settings::AppSettings;
+use speech::speaker::END_GRACE;
 use state::NostrRegistry;
 use ui::{
-    Avatar, MessageMedia, MessageRow, REACTION_EMOJIS, ReactionBar, ReplyPreview, group_reactions,
+    Avatar, MessageActions, MessageMedia, MessageRow, REACTION_EMOJIS, ReactionBar, ReplyPreview,
-    react_menu,
+    group_reactions, react_menu,
 };
 
 use crate::GroupPanel;
@@ -26,6 +28,8 @@ pub(crate) struct Actions {
     pub panel: WeakEntity<GroupPanel>,
     pub admin: bool,
     pub pinned: bool,
+    /// Whether a message is currently being spoken, disabling the speech buttons
+    pub speaking: bool,
     pub persons: Entity<Directory>,
 }
 
@@ -87,10 +91,44 @@ pub(crate) fn render(
     let admin = actions.admin;
     let pinned = actions.pinned;
     let group = actions.group;
-    let panel = actions.panel;
+    let speech_panel = actions.panel;
+    let panel = speech_panel.clone();
+    let speech_text = message.preview();
+    let speech_id = message.id;
+    let speaking = actions.speaking;
+
+    let row = row.overlay(
+        MessageActions::new().speech(speaking, move |_event, _window, cx| {
+            _ = speech_panel.update_in(cx, |panel, window, cx| {
+                match panel.speaker.speak(&speech_text) {
+                    Ok(duration) => {
+                        panel.speech_reveal = Some((speech_id, duration));
+
+                        let session = (speech_id, duration);
+                        let timer = cx.background_executor().timer(duration + END_GRACE);
+
+                        panel.reveal_timeout = Some(cx.spawn(async move |this, cx| {
+                            timer.await;
+                            this.update(cx, |this, cx| {
+                                if this.speech_reveal.as_ref() == Some(&session) {
+                                    this.speech_reveal = None;
+                                    cx.notify();
+                                }
+                            })?;
+
+                            Ok(())
+                        }));
+
+                        cx.notify();
+                    }
+                    Err(error) => {
+                        window.push_notification(Notification::error(error.to_string()), cx)
+                    }
+                }
+            });
+        }),
+    );
 
-    // Reading the panel here would re-enter its ongoing update, so resolve
-    // the current user's custom emojis through the shared person directory.
     let custom = nostr
         .read(cx)
         .current_user()

crates/speech/Cargo.toml

@@ -11,6 +11,7 @@ gpui.workspace = true
 gpui-kit = { workspace = true, features = ["speech"] }
 flume.workspace = true
 instant.workspace = true
+tts.workspace = true
 
 [target.'cfg(not(target_arch = "wasm32"))'.dependencies]
 common = { path = "../common" }

crates/speech/src/lib.rs

@@ -2,6 +2,7 @@
 pub mod model;
 #[cfg(not(target_arch = "wasm32"))]
 mod parakeet;
+pub mod speaker;
 #[cfg(target_arch = "wasm32")]
 mod web;
 

crates/speech/src/speaker.rs

@@ -0,0 +1,88 @@
+use std::time::Duration;
+
+use tts::Tts;
+
+/// Words per minute of an engine at its normal rate, for estimating speech duration.
+const NORMAL_WPM: f32 = 160.0;
+
+/// Extra time past the estimate before speech is considered finished, for backends that
+/// cannot report utterance end.
+pub const END_GRACE: Duration = Duration::from_secs(5);
+
+/// Reads text aloud through the platform's speech backend, created on first use.
+pub struct Speaker {
+    tts: Option<Tts>,
+    end_tx: flume::Sender<()>,
+    end_rx: flume::Receiver<()>,
+}
+
+impl Speaker {
+    pub fn new() -> Self {
+        let (end_tx, end_rx) = flume::unbounded();
+
+        Self {
+            tts: None,
+            end_tx,
+            end_rx,
+        }
+    }
+
+    /// Cloneable stream of utterance-finish events.
+    pub fn end_events(&self) -> flume::Receiver<()> {
+        self.end_rx.clone()
+    }
+
+    /// Speak the text, interrupting what this speaker is already saying.
+    pub fn speak(&mut self, text: &str) -> Result<Duration, tts::Error> {
+        let end_tx = self.end_tx.clone();
+        let tts = match self.tts.as_ref() {
+            Some(tts) => tts,
+            None => self.tts.insert({
+                let tts = Tts::default()?;
+                if let Err(error) = tts.on_utterance_end(move |_| {
+                    // A send error only means the panel dropped its event loop.
+                    let _ = end_tx.send(());
+                }) {
+                    log::warn!("utterance end events unavailable ({error})");
+                }
+                tts
+            }),
+        };
+        tts.speak(text, true)?;
+        Ok(Self::estimate(tts, text))
+    }
+
+    /// Stop speaking, a speaker that never spoke is already silent.
+    pub fn stop(&mut self) -> Result<(), tts::Error> {
+        match self.tts.as_ref() {
+            Some(tts) => tts.stop()?,
+            None => return Ok(()),
+        }
+
+        Ok(())
+    }
+
+    /// Best-effort speaking state, a speaker that never spoke is silent.
+    pub fn speaking(&self) -> bool {
+        self.tts
+            .as_ref()
+            .and_then(|tts| tts.is_speaking().ok())
+            .unwrap_or(false)
+    }
+
+    /// How long speaking the text is expected to take at the engine's current speed.
+    fn estimate(tts: &Tts, text: &str) -> Duration {
+        let words = text.split_whitespace().count().max(1) as f32;
+        let speed = match (tts.get_rate(), tts.normal_rate()) {
+            (Ok(rate), Ok(normal)) if normal > 0. => (rate / normal).clamp(0.2, 4.),
+            _ => 1.,
+        };
+        Duration::from_secs_f32(words / NORMAL_WPM * 60. / speed)
+    }
+}
+
+impl Default for Speaker {
+    fn default() -> Self {
+        Self::new()
+    }
+}

crates/ui/src/icon.rs

@@ -26,6 +26,7 @@ pub enum IconName {
     PlusChat,
     Profile,
     Relay,
+    Speech,
     Search,
     Settings,
     Settings2,
@@ -58,6 +59,7 @@ impl IconNamed for IconName {
             Self::PlusChat => "icons/plus-chat.svg",
             Self::Profile => "icons/profile.svg",
             Self::Relay => "icons/relay.svg",
+            Self::Speech => "icons/speech.svg",
             Self::Search => "icons/search.svg",
             Self::Settings => "icons/settings.svg",
             Self::Settings2 => "icons/settings2.svg",

crates/ui/src/message.rs

@@ -1,19 +1,21 @@
 use std::collections::BTreeMap;
 use std::rc::Rc;
+use std::time::Duration;
 
 use gpui::prelude::FluentBuilder as _;
 use gpui::{
-    AnyElement, App, ClickEvent, Context, ElementId, InteractiveElement as _, IntoElement,
+    Animation, AnimationExt as _, AnyElement, App, ClickEvent, Context, ElementId,
-    MouseButton, MouseDownEvent, ObjectFit, ParentElement as _, RenderOnce, SharedString,
+    InteractiveElement as _, IntoElement, MouseButton, MouseDownEvent, ObjectFit,
-    SharedUri, StatefulInteractiveElement as _, StyleRefinement, Styled, StyledImage, Window, div,
+    ParentElement as _, RenderOnce, SharedString, SharedUri, StatefulInteractiveElement as _,
-    img, px,
+    StyleRefinement, Styled, StyledImage, Window, div, img, px,
 };
 use gpui_kit::assets::IconName;
 use gpui_kit::component::button::{Button, ButtonVariants as _};
 use gpui_kit::component::clipboard::Clipboard;
 use gpui_kit::component::menu::{DropdownMenu as _, PopupMenu, PopupMenuItem};
 use gpui_kit::component::{
-    ActiveTheme, InteractiveElementExt as _, Sizable as _, StyledExt as _, h_flex, v_flex,
+    ActiveTheme, Disableable, InteractiveElementExt as _, Sizable as _, StyledExt as _, h_flex,
+    v_flex,
 };
 use smallvec::SmallVec;
 
@@ -597,13 +599,14 @@ pub fn react_menu(
     })
 }
 
-/// A hovered message's action bar: quick reactions, reply, copy, and an overflow menu.
+/// A hovered message's action bar: quick reactions, reply, copy, read aloud, and an overflow menu.
 #[derive(IntoElement)]
 pub struct MessageActions {
     style: StyleRefinement,
     reactions: SmallVec<[(&'static str, SharedClickListener); 4]>,
     reply: Option<SharedClickListener>,
     copy: Option<(ElementId, SharedString)>,
+    speech: Option<(bool, SharedClickListener)>,
     dropdown: Option<DropdownBuilder>,
 }
 
@@ -615,6 +618,7 @@ impl MessageActions {
             reactions: SmallVec::new(),
             reply: None,
             copy: None,
+            speech: None,
             dropdown: None,
         }
     }
@@ -642,6 +646,16 @@ impl MessageActions {
         self
     }
 
+    /// Add the read-aloud button, disabled while a message is already being spoken.
+    pub fn speech(
+        mut self,
+        disabled: bool,
+        handler: impl Fn(&ClickEvent, &mut Window, &mut App) + 'static,
+    ) -> Self {
+        self.speech = Some((disabled, Rc::new(handler)));
+        self
+    }
+
     /// Fill the bar's overflow menu.
     pub fn dropdown(
         mut self,
@@ -671,10 +685,13 @@ impl RenderOnce for MessageActions {
             reactions,
             reply,
             copy,
+            speech,
             dropdown,
         } = self;
 
-        let has_center = reply.is_some() || copy.is_some();
+        // The leading separator only separates reactions from the center buttons
+        let has_center =
+            !reactions.is_empty() && (reply.is_some() || copy.is_some() || speech.is_some());
 
         h_flex()
             .p_0p5()
@@ -712,6 +729,17 @@ impl RenderOnce for MessageActions {
             .when_some(copy, |this, (id, value)| {
                 this.child(Clipboard::new(id).value(value).tooltip("Copy").small())
             })
+            .when_some(speech, |this, (disabled, speech)| {
+                this.child(
+                    Button::new("speech")
+                        .icon(crate::IconName::Speech)
+                        .tooltip("Read aloud")
+                        .small()
+                        .ghost()
+                        .disabled(disabled)
+                        .on_click(move |event, window, cx| speech(event, window, cx)),
+                )
+            })
             .when_some(dropdown, |this, dropdown| {
                 this.child(div().flex_shrink_0().h_4().w_px().bg(cx.theme().border))
                     .child(
@@ -726,3 +754,44 @@ impl RenderOnce for MessageActions {
             .refine_style(&style)
     }
 }
+
+/// A message's text revealed word by word, paced by the estimated speech duration.
+#[derive(IntoElement)]
+pub struct RevealText {
+    text: SharedString,
+    duration: Duration,
+}
+
+impl RevealText {
+    pub fn new(text: impl Into<SharedString>, duration: Duration) -> Self {
+        Self {
+            text: text.into(),
+            duration,
+        }
+    }
+}
+
+impl RenderOnce for RevealText {
+    fn render(self, _window: &mut Window, _cx: &mut App) -> impl IntoElement {
+        let words: Vec<SharedString> = self
+            .text
+            .split_whitespace()
+            .map(SharedString::from)
+            .collect();
+        let total = words.len().max(1) as f32;
+
+        div().flex().flex_wrap().gap_x_1().with_animation(
+            "speech-reveal",
+            Animation::new(self.duration).with_max_fps(30.),
+            move |this, progress| {
+                let visible = ((progress * total).ceil() as usize).min(words.len());
+
+                this.children(words.iter().enumerate().map(|(ix, word)| {
+                    div()
+                        .when(ix >= visible, |this| this.invisible())
+                        .child(word.clone())
+                }))
+            },
+        )
+    }
+}