@tychenjiajun/pi-voice 0.7.0 → 0.8.1

package/src/languages.ts

@@ -3,9 +3,10 @@ export function canonicalLanguage(language: string): string {
   return language.trim().toLowerCase().split("-", 1)[0] ?? "";
 }
 
-// Model cards and benchmarks use different ISO codes for these languages.
-// This identity is for matching, never a code to send directly to a backend.
-const LANGUAGE_ALIASES: Readonly<Record<string, string>> = { tl: "fil", no: "nb" };
+// Model cards, benchmarks, and the language detector use different ISO codes
+// for these languages (iw/jw are legacy codes for Hebrew and Javanese). This
+// identity is for matching, never a code to send directly to a backend.
+const LANGUAGE_ALIASES: Readonly<Record<string, string>> = { tl: "fil", no: "nb", iw: "he", jw: "jv" };
 
 export function languageIdentity(language: string): string {
   const base = canonicalLanguage(language);

package/src/model-activation.ts

@@ -1,13 +1,17 @@
-import type { CatalogModel } from "./catalog.js";
-import { downloadCatalogModel, findCachedCatalogModel, type CachedCatalogModel } from "./models.js";
+import { LANGUAGE_DETECTION_MODEL, displayLanguage, type CatalogModel } from "./catalog.js";
+import { settingsLanguagePolicy } from "./language-detection-policy.js";
+import { describeError, log } from "./log.js";
+import { downloadCatalogModel, findCachedCatalogModel, pendingDownloads, type CachedCatalogModel } from "./models.js";
 import { writeSettings, type TranscribeSettings } from "./settings.js";
 
+export type ActivationProgress = { downloaded: number; total: number; message?: string };
+
 export type CatalogModelActivation = (
   model: CatalogModel,
   options: {
     cached: CachedCatalogModel | undefined;
     signal: AbortSignal;
-    onProgress: (progress: { downloaded: number; total: number }) => void;
+    onProgress: (progress: ActivationProgress) => void;
   },
 ) => Promise<{ path: string }>;
 
@@ -20,6 +24,7 @@ export type CatalogModelActivation = (
 export function createModelActivation(options: {
   buildSettings: (model: CatalogModel, path: string) => TranscribeSettings;
   onCommitted: (settings: TranscribeSettings) => void;
+  onWarning: (message: string) => void;
 }): {
   activate: CatalogModelActivation;
   /** Resolves once every commit enqueued so far has landed (or failed). */
@@ -32,6 +37,11 @@ export function createModelActivation(options: {
     { cached, signal, onProgress },
   ) => {
     let path: string;
+    let detectorFailed = false;
+    const pending = pendingDownloads(model, Boolean(cached));
+    const detectorMissing = pending.includes(LANGUAGE_DETECTION_MODEL);
+    const asrBytes = cached ? 0 : model.size;
+    const total = pending.reduce((sum, item) => sum + item.size, 0);
     if (cached) {
       // The picker listed the cache when it opened; re-check so settings
       // never point at a file that has since been evicted. Integrity is
@@ -44,9 +54,33 @@ export function createModelActivation(options: {
       }
       path = stillCached.path;
     } else {
-      path = await downloadCatalogModel(model, { signal, onProgress });
+      path = await downloadCatalogModel(model, {
+        signal,
+        onProgress: (progress) => onProgress({ ...progress, total }),
+      });
     }
 
+    if (detectorMissing) {
+      try {
+        onProgress({ downloaded: asrBytes, total, message: "Downloading language detector…" });
+        await downloadCatalogModel(LANGUAGE_DETECTION_MODEL, {
+          signal,
+          onProgress: (progress) => onProgress({
+            downloaded: asrBytes + progress.downloaded, total,
+            message: "Downloading language detector…",
+          }),
+        });
+      } catch (error) {
+        // Cancellation is never a settings commit. Other failures still
+        // select the model: without the detector, Auto falls back to the
+        // first preferred language, and the first take says how to fix it.
+        signal.throwIfAborted();
+        log.warn(`Language detector download failed: ${describeError(error)}`);
+        detectorFailed = true;
+      }
+    }
+
+    let committed: TranscribeSettings | undefined;
     const commit = commitQueue.then(async () => {
       // Skip the write when this selection was superseded or its download
       // was cancelled while the commit sat in the queue.
@@ -60,6 +94,7 @@ export function createModelActivation(options: {
           `Could not save settings: ${error instanceof Error ? error.message : String(error)}`,
         );
       }
+      committed = settings;
       options.onCommitted(settings);
     });
     commitQueue = commit.then(
@@ -67,6 +102,17 @@ export function createModelActivation(options: {
       () => undefined,
     );
     await commit;
+    // Only Auto detect uses the detector, so other settings need no warning.
+    const policy = committed && settingsLanguagePolicy(committed);
+    if (detectorFailed && policy?.kind === "detect") {
+      try {
+        options.onWarning(
+          `Selected ${model.name}, but the language detector didn't download. ` +
+          `Auto detect will use ${displayLanguage(policy.languages[0]!)} until it does. ` +
+          "To retry the download, choose Auto detect in /voice-settings.",
+        );
+      } catch { /* Notification is not part of the save. */ }
+    }
     return { path };
   };
 

package/src/model-cells.ts

@@ -16,6 +16,8 @@ import {
   type AccuracyLetter,
   type RecommendationRole,
 } from "./recommendations.js";
+import { requiresLanguageDetector, undetectablePreferences } from "./language-detection-policy.js";
+import { findIncompleteDownload, pendingDownloads } from "./models.js";
 import { LIST_PADDING, padToWidth, selectionMarker } from "./ui-components.js";
 
 type UiTheme = ExtensionContext["ui"]["theme"];
@@ -181,18 +183,16 @@ export function modelDetailText(
   width: number,
   padding: number,
   feedback?: { type: "success" | "error" | "muted"; text: string },
-  showManualSelection = false,
 ): string {
   const canonicalLanguages = [...new Set(model.languages.map(canonicalLanguage))];
   const features = [
     canonicalLanguages.length === 1
       ? `${displayLanguage(canonicalLanguages[0]!)} only`
       : `${canonicalLanguages.length} languages`,
-    model.capabilities.languageDetection
-      ? "auto language detection"
-      : showManualSelection
-        ? "manual language selection"
-        : undefined,
+    // Native detection, or the shared detector for multilingual models.
+    model.capabilities.languageDetection || requiresLanguageDetector(model)
+      ? "automatic language detection"
+      : undefined,
   ].filter((value): value is string => Boolean(value));
   const description = truncateToWidth(
     model.description,
@@ -205,6 +205,28 @@ export function modelDetailText(
   return `${description}\n${theme.fg("dim", features.join(" · "))}${feedbackText}`;
 }
 
+/** A picker's action for a model: "choose", or what is left to download. */
+export function downloadAction(model: CatalogModel, modelCached: boolean): string {
+  const pending = pendingDownloads(model, modelCached);
+  if (!pending.length) return "choose";
+  const partial = pending.map(findIncompleteDownload);
+  const bytes = pending.reduce((sum, item, index) => sum + item.size - (partial[index]?.bytes ?? 0), 0);
+  return partial.some(Boolean)
+    ? `resume download (${formatBinarySize(bytes)} left)`
+    : `download ${formatBinarySize(bytes)}`;
+}
+
+/** Why Auto detect is missing for these preferences, if a language is undetectable. */
+export function autoUnavailableNotice(
+  model: CatalogModel,
+  preferredLanguages: readonly string[],
+): string | undefined {
+  const skipped = undetectablePreferences(model, preferredLanguages).map(displayLanguage);
+  if (!skipped.length) return undefined;
+  const languages = new Intl.ListFormat("en", { type: "disjunction" }).format(skipped);
+  return `Auto detect is unavailable for your preferred languages because it can't detect ${languages}.`;
+}
+
 /**
  * The catalog's search matching, shared by every page that lists models.
  * Each word of the query must appear literally in the model's id, name,

package/src/model-picker.ts

@@ -31,6 +31,8 @@ import {
   type ModelBenchmark,
 } from "./recommendations.js";
 import {
+  autoUnavailableNotice,
+  downloadAction,
   MANUAL_LANGUAGE_TAG,
   ON_DISK_LABEL,
   modelDetailText,
@@ -41,7 +43,7 @@ import {
 import { ModelSelectionController } from "./model-selection-controller.js";
 import { ModelRatingsHelp } from "./model-ratings-help.js";
 import type { CatalogModelActivation } from "./model-activation.js";
-import { findIncompleteDownload } from "./models.js";
+import { autoLanguagePolicy } from "./language-detection-policy.js";
 import { VoiceKeys } from "./keybindings.js";
 import type { TranscriptionLanguage } from "./settings.js";
 import {
@@ -656,8 +658,11 @@ export class CatalogModelPicker extends Container implements Focusable {
     const feedbackLines = this.selection.feedback
       ? new Text(this.selection.feedback.text, TEXT_PADDING, 0).render(width).length
       : 0;
-    // One description line plus the features line.
-    return 2 + feedbackLines;
+    const model = this.highlightedModel();
+    const notices = model ? this.modelNotices(model) : "";
+    const noticeLines = notices ? new Text(notices, TEXT_PADDING, 0).render(width).length : 0;
+    // Description, features, and the undetectable-language warning.
+    return 2 + feedbackLines + noticeLines;
   }
 
   private refresh(): void {
@@ -783,12 +788,17 @@ export class CatalogModelPicker extends Container implements Focusable {
           this.renderWidth,
           TEXT_PADDING,
           this.selection.feedback,
-        ),
+        ) + this.modelNotices(selected),
       );
     }
     this.finishFooter(query);
   }
 
+  private modelNotices(model: CatalogModel): string {
+    const notice = autoUnavailableNotice(model, this.preferredLanguages);
+    return notice ? `\n${this.theme.fg("warning", notice)}` : "";
+  }
+
   private finishFooter(query: string): void {
     const displayedId = this.selection.displayedModelId;
     const total = CATALOG_MODELS.length;
@@ -803,10 +813,8 @@ export class CatalogModelPicker extends Container implements Focusable {
     const highlighted = this.highlightedModel();
     const confirmLabel = this.rows[this.selectedIndex]?.type === "fold"
       ? this.folded ? "show" : "hide"
-      : highlighted && !this.selection.cachedById.has(highlighted.id)
-        ? findIncompleteDownload(highlighted)
-          ? "resume download"
-          : `download ${formatBinarySize(highlighted.size)}`
+      : highlighted
+        ? downloadAction(highlighted, this.selection.cachedById.has(highlighted.id))
         : "choose";
     this.footer.setText(
       `${this.theme.fg("dim", shown)}  ${statusLegend}  ${this.keys.hint("voice.models.ratingsHelp", "rating guide")}\n${this.keys.navHint("navigate")}  ${this.keys.hint("tui.select.confirm", confirmLabel)}  ${this.keys.hint("tui.select.cancel", closeLabel)}`,
@@ -1003,10 +1011,19 @@ export function createTranscriptionLanguagePicker(
         left.label.localeCompare(right.label) ||
         left.value.localeCompare(right.value),
     );
+  const policy = autoLanguagePolicy(model, preferredLanguages);
+  const autoAvailable = policy.kind !== "unsupported";
+  const auto: SingleSelectChoice<TranscriptionLanguage> = policy.kind === "detect"
+    ? {
+      value: "auto",
+      label: "Auto detect",
+      description: `Detects which of your preferred languages you're speaking (${new Intl.ListFormat("en", { type: "disjunction" }).format(
+        policy.languages.map((language) => transcriptionLanguageName(language, model.languages)),
+      )}) in each recording.`,
+    }
+    : { value: "auto", label: "Auto detect" };
   const choices: SingleSelectChoice<TranscriptionLanguage>[] = [
-    ...(model.capabilities.languageDetection
-      ? [{ value: "auto", label: "Auto detect" }]
-      : []),
+    ...(autoAvailable ? [auto] : []),
     ...languages,
   ];
   return new SingleSelectPicker(
@@ -1017,7 +1034,7 @@ export function createTranscriptionLanguagePicker(
     current,
     {
       title: "Choose transcription language",
-      subtitle: model.capabilities.languageDetection
+      subtitle: autoAvailable
         ? "Language expected in recordings, or automatic detection."
         : "Language expected in recordings.",
       searchable: true,

package/src/model-selection-controller.ts

@@ -1,6 +1,6 @@
 import type { CatalogModel } from "./catalog.js";
 import type { CatalogModelActivation } from "./model-activation.js";
-import { findCachedCatalogModel, type CachedCatalogModel } from "./models.js";
+import { findCachedCatalogModel, pendingDownloads, type CachedCatalogModel } from "./models.js";
 
 export type DownloadState = {
   readonly model: CatalogModel;
@@ -94,7 +94,8 @@ export class ModelSelectionController<R> {
   select(model: CatalogModel): void {
     if (!this.acceptsInput || this.download) return;
     const cached = this.cachedById.get(model.id);
-    if (!this.target && cached && model.id === this.committedModelId &&
+    const downloads = pendingDownloads(model, Boolean(cached)).length > 0;
+    if (!downloads && !this.target && model.id === this.committedModelId &&
         (this.selectedDuringSession || !this.options.advance)) {
       this.selectedDuringSession = true;
       if (this.options.advance) this.finish(this.options.completion);
@@ -107,9 +108,9 @@ export class ModelSelectionController<R> {
     this.target = target;
     this.feedback = undefined;
     this.samples = [];
-    this.download = cached ? undefined : {
+    this.download = downloads ? {
       model, downloaded: 0, total: 0, message: "Connecting to Hugging Face…",
-    };
+    } : undefined;
     this.changed();
 
     // Catch synchronous implementations too. The real pipeline reports only
@@ -119,15 +120,15 @@ export class ModelSelectionController<R> {
       work = this.activate(model, {
         cached,
         signal: target.controller.signal,
-        onProgress: ({ downloaded, total }) => {
+        onProgress: ({ downloaded, total, message }) => {
           if (this.disposed || this.closed || this.target !== target || target.controller.signal.aborted) return;
           if (!this.download) return;
           const firstReport = this.download.total === 0 && total > 0;
           this.download = {
             model, downloaded, total,
-            message: firstReport
+            message: message ?? (firstReport
               ? downloaded > 0 ? "Resuming download from Hugging Face…" : "Downloading from Hugging Face…"
-              : this.download.message,
+              : this.download.message),
           };
           this.samples.push({ time: this.now(), bytes: downloaded });
           if (this.samples.length > 64) this.samples.shift();
@@ -147,9 +148,13 @@ export class ModelSelectionController<R> {
         if (this.target !== target) { this.changed(); return; }
         this.target = undefined;
         this.download = undefined;
-        this.feedback = cached ? undefined : {
+        // A failed detector download still selects the model; its own warning explains.
+        const detectorDownloaded = cached && !pendingDownloads(model, true).length;
+        this.feedback = !cached ? {
           type: "success", text: `✓ Downloaded and selected ${model.name}`,
-        };
+        } : detectorDownloaded && downloads ? {
+          type: "success", text: `✓ Selected ${model.name} and downloaded the language detector`,
+        } : undefined;
         if (this.pendingExit) this.finish(this.pendingExit.result);
         else if (this.options.advance) this.finish(this.options.completion);
         else this.changed();

package/src/models.ts

@@ -21,7 +21,8 @@ import {
 } from "node:fs/promises";
 import { dirname, join, relative } from "node:path";
 import type { ReadableStream } from "node:stream/web";
-import type { CatalogModel } from "./catalog.js";
+import { LANGUAGE_DETECTION_MODEL, type CatalogModel } from "./catalog.js";
+import { requiresLanguageDetector } from "./language-detection-policy.js";
 
 function repositoryCacheDirectory(model: CatalogModel): string {
   return join(
@@ -261,3 +262,16 @@ export function findIncompleteDownload(
   }
   return undefined;
 }
+
+/**
+ * What choosing `model` still has to download: the model unless it is cached,
+ * plus the shared language detector when the model relies on it and it is
+ * not cached either.
+ */
+export function pendingDownloads(model: CatalogModel, modelCached: boolean): CatalogModel[] {
+  const pending = modelCached ? [] : [model];
+  if (requiresLanguageDetector(model) && !findCachedCatalogModel(LANGUAGE_DETECTION_MODEL)) {
+    pending.push(LANGUAGE_DETECTION_MODEL);
+  }
+  return pending;
+}

package/src/onboarding.ts

@@ -21,6 +21,7 @@ import {
   DEFAULT_MICROPHONE,
   settingsForModel,
   settingsRemoteOnly,
+  settingsWithPreferredLanguages,
   writeSettings,
   type BackendSettings,
   type ChineseOutput,
@@ -47,10 +48,12 @@ type SettingsOptions = NonNullable<Parameters<typeof settingsForModel>[2]>;
 function createSettingsActivation(
   options: () => SettingsOptions,
   onCommitted: (settings: TranscribeSettings) => void,
+  ctx: ExtensionContext,
 ) {
   return createModelActivation({
     buildSettings: (model, path) => settingsForModel(model.id, path, options()),
     onCommitted,
+    onWarning: (message) => ctx.ui.notify(message, "warning"),
   });
 }
 
@@ -67,6 +70,8 @@ type ModelSelectionOptions = {
   polishEnabled?: boolean;
   /** Preserved across the model re-pick; absent means the agent's current model. */
   polishModel?: PolishModelRef;
+  voiceActivityDetection?: boolean;
+  longformChunking?: boolean;
   /** Persists language changes made before this flow activates a model. */
   onPreferredLanguagesChange?: (languages: string[]) => Promise<void>;
   postActivation?: CatalogModelPostActivation;
@@ -94,11 +99,14 @@ export async function runModelSelection(
       backend: configured?.backend ?? options.backend,
       polishEnabled: configured?.polishEnabled ?? options.polishEnabled ?? false,
       polishModel: configured?.polishModel ?? options.polishModel,
+      voiceActivityDetection: configured?.voiceActivityDetection ?? options.voiceActivityDetection,
+      longformChunking: configured?.longformChunking ?? options.longformChunking,
     }),
     (settings) => {
       configured = settings;
       currentModelId = settings.model?.id;
     },
+    ctx,
   );
 
   // One pane switches between the models on disk and downloads new ones.
@@ -123,7 +131,7 @@ export async function runModelSelection(
     if (changed) {
       try {
         if (configured) {
-          const updated = { ...configured, preferredLanguages: changed.languages };
+          const updated = settingsWithPreferredLanguages(configured, changed.languages);
           await writeSettings(updated);
           configured = updated;
         } else {
@@ -225,10 +233,13 @@ export async function changeOnboardingModel(
       backend: current.backend,
       polishEnabled: current.polishEnabled,
       polishModel: current.polishModel,
+      voiceActivityDetection: current.voiceActivityDetection,
+      longformChunking: current.longformChunking,
     }),
     (settings) => {
       chosen = settings;
     },
+    ctx,
   );
 
   while (true) {
@@ -356,6 +367,7 @@ export async function runOnboarding(
       (settings) => {
         configured = settings;
       },
+      ctx,
     );
 
     let changeLanguages = false;

package/src/pcm-chunker.ts

@@ -2,13 +2,18 @@ import { CAPTURE_SAMPLE_RATE } from "./audio-constants.js";
 
 const STREAM_CHUNK_SAMPLES = CAPTURE_SAMPLE_RATE / 2;
 
-/** Coalesces recorder chunks into fresh Float32 PCM chunks without timers. */
+/**
+ * Coalesces recorder chunks into fresh Float32 PCM chunks without timers,
+ * along with their Silero frame probabilities: a chunk has them only when
+ * every piece of it did.
+ */
 export class PcmChunker {
   private pieces: Float32Array[] = [];
+  private probs: number[] | undefined = [];
   private sampleCount = 0;
 
   constructor(
-    private readonly onChunk: (chunk: Float32Array) => void,
+    private readonly onChunk: (chunk: Float32Array, probs?: readonly number[]) => void,
     private readonly targetSamples = STREAM_CHUNK_SAMPLES,
   ) {
     if (!Number.isInteger(targetSamples) || targetSamples <= 0) {
@@ -16,9 +21,11 @@ export class PcmChunker {
     }
   }
 
-  push(samples: Float32Array): void {
+  push(samples: Float32Array, probs?: readonly number[]): void {
     if (samples.length === 0) return;
     this.pieces.push(samples);
+    if (probs && this.probs) this.probs.push(...probs);
+    else this.probs = undefined;
     this.sampleCount += samples.length;
     if (this.sampleCount >= this.targetSamples) this.emit();
   }
@@ -29,6 +36,7 @@ export class PcmChunker {
 
   discard(): void {
     this.pieces = [];
+    this.probs = [];
     this.sampleCount = 0;
   }
 
@@ -39,8 +47,10 @@ export class PcmChunker {
       pcm.set(piece, offset);
       offset += piece.length;
     }
+    const probs = this.probs;
     this.pieces = [];
+    this.probs = [];
     this.sampleCount = 0;
-    this.onChunk(pcm);
+    this.onChunk(pcm, probs);
   }
 }

package/src/recommendation-picker.ts

@@ -13,6 +13,8 @@ import {
   formatBinarySize,
 } from "./catalog.js";
 import type { CatalogModelActivation } from "./model-activation.js";
+import { undetectablePreferences } from "./language-detection-policy.js";
+import { autoUnavailableNotice, downloadAction } from "./model-cells.js";
 import { ModelSelectionController } from "./model-selection-controller.js";
 import { VoiceKeys } from "./keybindings.js";
 import {
@@ -202,9 +204,10 @@ export class RecommendedModelPicker extends Container implements Focusable {
       } else {
         text = "A good balance of speed and accuracy.";
       }
-      if (this.languages.length > 1 && recommendation.model.capabilities.languageDetection) {
-        const scope = this.languages.length === 2 ? "both languages" : "all your languages";
-        text += ` It switches between ${scope} automatically.`;
+      // Recommended models support every chosen language, so Auto covers
+      // them all unless the detector cannot identify one.
+      if (this.languages.length > 1 && undetectablePreferences(recommendation.model, this.languages).length === 0) {
+        text += " It detects which language you're speaking automatically.";
       }
       return text;
     }
@@ -272,9 +275,8 @@ export class RecommendedModelPicker extends Container implements Focusable {
         ),
       );
     }
-    if (this.languages.length > 1 && !recommendation.model.capabilities.languageDetection) {
-      details.addChild(new Text(this.theme.fg("warning", "You will need to change the transcription language manually."), LIST_PADDING + 2, 0));
-    }
+    const notice = autoUnavailableNotice(recommendation.model, this.languages);
+    if (notice) details.addChild(new Text(this.theme.fg("warning", notice), LIST_PADDING + 2, 0));
     return details;
   }
 
@@ -387,7 +389,7 @@ export class RecommendedModelPicker extends Container implements Focusable {
     if (!row) return "choose";
     if (row.type !== "model") return this.rowLabel(row).action;
     const model = row.recommendation.model;
-    return this.selection.cachedById.has(model.id) ? "choose" : `download ${formatBinarySize(model.size)}`;
+    return downloadAction(model, this.selection.cachedById.has(model.id));
   }
 
   override render(width: number): string[] {

package/src/recommendations.ts

@@ -7,6 +7,8 @@ import {
   type CatalogModel,
 } from "./catalog.js";
 
+import { undetectablePreferences } from "./language-detection-policy.js";
+
 export type MachineTier = "accelerated" | "cpu";
 
 export type RecommendationRole = "best" | "fast" | "accurate";
@@ -211,7 +213,7 @@ function scoreModels(
       worstCiLower: accuracyCells[worstIndex]!.ciLower,
       waitSeconds,
       cpuWaitSeconds: cpuXrt ? method.dictationSeconds / cpuXrt : undefined,
-      manual: wanted.length > 1 && !model.capabilities.languageDetection,
+      manual: wanted.length > 1 && undetectablePreferences(model, languages).length > 0,
       score: error + method.waitWeight * Math.max(0, waitSeconds - method.freeWaitSeconds),
     });
   }
@@ -284,7 +286,7 @@ function fallbackRecommendation(
     (model) =>
       wanted.every((language) =>
         modelMatchesRecommendationLanguage(model, language),
-      ) && (wanted.length === 1 || model.capabilities.languageDetection),
+      ) && (wanted.length === 1 || undetectablePreferences(model, languages).length === 0),
   );
   const model = rankCatalogModels(compatible.length ? compatible : models, wanted)[0];
   return model

package/src/runtime.ts

@@ -15,6 +15,7 @@ import { polishTranscript } from "./polish.js";
 import type { TranscribeSettings } from "./settings.js";
 import { displayShortcut } from "./shortcut-core.js";
 import { TranscriptionService } from "./transcription-service.js";
+import { VoiceActivityDetection } from "./vad.js";
 import type { RecordingMeter } from "./visualizer.js";
 
 type ActiveRecording = {
@@ -67,7 +68,8 @@ export function createPiVoiceRuntime(
   let settingsWarningShown = false;
   let audioModulePromise: Promise<typeof import("./audio.js")> | undefined;
   let visualizerModulePromise: Promise<typeof import("./visualizer.js")> | undefined;
-  const transcriptionService = new TranscriptionService();
+  const voiceActivity = new VoiceActivityDetection();
+  const transcriptionService = new TranscriptionService(undefined, (pcm, signal) => voiceActivity.scoreClip(pcm, signal));
 
   function loadAudio(): Promise<typeof import("./audio.js")> {
     return (audioModulePromise ??= import("./audio.js"));
@@ -150,6 +152,8 @@ export function createPiVoiceRuntime(
       backend: previous.backend,
       polishEnabled: previous.polishEnabled,
       polishModel: previous.polishModel,
+      voiceActivityDetection: previous.voiceActivityDetection,
+      longformChunking: previous.longformChunking,
       postActivation: "advance",
     });
     if (configured) rememberSettings(configured);
@@ -294,6 +298,7 @@ export function createPiVoiceRuntime(
     try {
       const result = await active.dictation.stop();
       if (shuttingDown) return;
+      if (result?.warning) ctx.ui.notify(result.warning, "warning");
       if (!result) {
         await reportDictationError(ctx, active.dictation);
       } else if (result.text) {
@@ -355,6 +360,7 @@ export function createPiVoiceRuntime(
       openCapture: openMicrophone,
       onAudio: (chunk) => meter.push(chunk),
       onChange: updateMeter,
+      createSpeechFilter: (settings, onSpeech) => voiceActivity.createFilter(settings, onSpeech),
     });
     function updateMeter(): void {
       meter.update({ modelState: controller.modelState, microphoneStopped: controller.interruption !== undefined });
@@ -514,6 +520,8 @@ export function createPiVoiceRuntime(
       operation?.catch(() => undefined),
       transcriptionService.shutdown().catch(() => undefined),
     ]);
+    // After the dictation, so no recording still holds a VAD session.
+    await voiceActivity.dispose().catch(() => undefined);
     recording = undefined;
     dictation = undefined;
     if (visualizerModulePromise) {