VoiceMessageTranscriber: Added Embed, model size estimates (#1302)

* feat(voiceMessageTranscriber): add embed and maintain horizontal options with model size indicators

* Apply suggestions from code review

Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>

* fix(voiceMessageTranscriber): scope parent selectors and resolve transcription race condition

* refactor(voiceMessageTranscriber): adhere strictly to .rules types and formatting

* fix(voiceMessageTranscriber): reliably wrap parent container for embed layout

* feat(voiceMessageTranscriber): add tt to authors

* some fixes

---------

Co-authored-by: thororen <78185467+thororen1234@users.noreply.github.com>
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>

src/equicordplugins/voiceMessageTranscriber.desktop/index.tsx

@@ -10,27 +10,88 @@ import { DataStore } from "@api/index";
 import { definePluginSettings } from "@api/Settings";
 import { BaseText } from "@components/BaseText";
 import { Button } from "@components/Button";
+import ErrorBoundary from "@components/ErrorBoundary";
 import { Flex } from "@components/Flex";
-import { FormSwitch } from "@components/FormSwitch";
 import { Heading } from "@components/Heading";
+import { ClockIcon, CopyIcon } from "@components/Icons";
 import { Span } from "@components/Span";
 import { copyToClipboard } from "@utils/clipboard";
-import { Devs } from "@utils/constants";
+import { Devs, EquicordDevs } from "@utils/constants";
+import { classes } from "@utils/misc";
 import definePlugin, { OptionType, PluginNative } from "@utils/types";
 import { RenderModalProps } from "@vencord/discord-types";
 import { findComponentByCodeLazy } from "@webpack";
-import { lodash, Modal, openModal, ScrollerAuto, SearchableSelect, useEffect, useRef, useState } from "@webpack/common";
+import { Modal, openModal, ScrollerAuto, SearchableSelect, Tooltip, useEffect, useRef, useState } from "@webpack/common";
 
-import { cl, decodeAudio, LANGUAGES, TranscriptionWorker } from "./utils";
+import { cl, decodeAudio, LANGUAGES, TranscriptionResult, TranscriptionWorker } from "./utils";
 const Native = VencordNative.pluginHelpers.VoiceMessageTranscriber as PluginNative<typeof import("./native")>;
 
 const ChannelListIcon = findComponentByCodeLazy("1-1-1ZM2 8a1");
-let ManaBaseRadioGroup;
+let ManaBaseRadioGroup: React.ComponentType<{ options: { name: string; value: string; }[]; value: string; onChange: (v: string) => void; }>;
+
+function formatTimestamp(seconds: number) {
+    const m = Math.floor(seconds / 60);
+    const s = Math.floor(seconds % 60);
+    return `${m.toString().padStart(2, "0")}:${s.toString().padStart(2, "0")}`;
+}
+
+function CloseIcon({ size = 14 }: { size?: number; }) {
+    return (
+        <svg width={size} height={size} viewBox="0 0 24 24" fill="none" stroke="currentColor" strokeWidth="2" strokeLinecap="round" strokeLinejoin="round">
+            <line x1="18" y1="6" x2="6" y2="18" />
+            <line x1="6" y1="6" x2="18" y2="18" />
+        </svg>
+    );
+}
+
+function CheckmarkIcon({ size = 14 }: { size?: number; }) {
+    return (
+        <svg width={size} height={size} viewBox="0 0 24 24" fill="none" stroke="currentColor" strokeWidth="2.5" strokeLinecap="round" strokeLinejoin="round">
+            <polyline points="20 6 9 17 4 12" />
+        </svg>
+    );
+}
+
+const MODEL_SIZES: Record<string, { quantized: string; full: string; }> = {
+    "Xenova/whisper-tiny": { quantized: "~40 MB", full: "~150 MB" },
+    "Xenova/whisper-base": { quantized: "~77 MB", full: "~290 MB" },
+    "Xenova/whisper-small": { quantized: "~250 MB", full: "~960 MB" },
+    "Xenova/whisper-medium": { quantized: "~765 MB", full: "~3.1 GB" },
+};
+
+function renderModelOption(option?: { label: string; value: string; }) {
+    if (!option) return null;
+    const isQuantized = settings.use(["quantized"])?.quantized ?? true;
+    const size = MODEL_SIZES[option.value]?.[isQuantized ? "quantized" : "full"];
+
+    return (
+        <div className={cl("model-option")}>
+            <span>{option.label}</span>
+            {size && (
+                <span className={cl("model-size")}>
+                    {size}
+                </span>
+            )}
+        </div>
+    );
+}
 
 const settings = definePluginSettings({
+    embed: {
+        type: OptionType.BOOLEAN,
+        description: "Display transcription directly in the voice message attachment instead of a modal.",
+        default: false,
+        restartNeeded: false
+    },
+    maintainHorizontal: {
+        type: OptionType.BOOLEAN,
+        description: "Maintain horizontal size for the embedded transcription box and expand vertically.",
+        default: false,
+        restartNeeded: false
+    },
     selectedModel: {
         type: OptionType.SELECT,
-        description: "Model size",
+        description: "Model size.",
         options: [
             {
                 label: "Tiny (Fastest, lowest accuracy)",
@@ -50,38 +111,45 @@ const settings = definePluginSettings({
                 value: "Xenova/whisper-medium"
             }
         ],
+        componentProps: {
+            renderOptionLabel: (option: { label: string; value: string; }) => renderModelOption(option),
+            renderOptionValue: (options: { label: string; value: string; }[]) => renderModelOption(options?.[0]),
+        },
         restartNeeded: false
     },
     quantized: {
         type: OptionType.BOOLEAN,
-        description: "Use quantized model (smaller size, slightly lower accuracy)",
+        description: "Use quantized models (smaller size, slight quality loss).",
         default: true,
         restartNeeded: false
     },
-    delete: {
+    deleteModalFiles: {
         type: OptionType.COMPONENT,
+        description: "Delete cached files from storage.",
         component: () => {
             const [size, setSize] = useState(0);
             const [deleteKeys, setDeleteKeys] = useState<string[]>([]);
 
             useEffect(() => {
+                let unmounted = false;
                 DataStore.entries().then(entries => {
-                    let size = 0;
-                    const keys = [] as string[];
-
-                    entries.forEach(([key, val]) => {
-                        if (typeof key === "string" && key.startsWith("VoiceMessageTranscriber") && lodash.isArrayBuffer(val)) {
+                    if (unmounted) return;
+                    let totalSize = 0;
+                    const keys: string[] = [];
+                    for (const [key, value] of entries) {
+                        if (typeof key === "string" && (key.startsWith("whisper-") || key.startsWith("onnx-"))) {
+                            totalSize += (value as string).length;
                             keys.push(key);
-                            size += val.byteLength ?? 0;
                         }
-                    });
-
-                    setSize(size);
+                    }
+                    setSize(totalSize);
                     setDeleteKeys(keys);
                 });
+                return () => { unmounted = true; };
             }, []);
 
             return <Button
+                disabled={size === 0}
                 variant="dangerPrimary"
                 onClick={() => {
                     DataStore.delMany(deleteKeys).then(() => { setSize(0); setDeleteKeys([]); });
@@ -166,7 +234,7 @@ function LanguageSelectionModal(props: { modalProps: RenderModalProps, src: stri
 function TranscriptionModal(props: { modalProps: RenderModalProps, src: string, options: { language: string, task: string; }; }) {
     const { modalProps, src, options } = props;
     const [status, setStatus] = useState<string>("initializing");
-    const [result, setResult] = useState<{ text: string, chunks: { timestamp: [number, number], text: string; }[]; } | null>(null);
+    const [result, setResult] = useState<TranscriptionResult | null>(null);
     const [showTimestamps, setShowTimestamps] = useState(false);
     const [error, setError] = useState<string | null>(null);
     const [copied, setCopied] = useState(false);
@@ -183,8 +251,8 @@ function TranscriptionModal(props: { modalProps: RenderModalProps, src: string,
 
                 let blob: Blob;
                 if (IS_DISCORD_DESKTOP || IS_EQUIBOP) {
-                    const arrayBuffer = await Native.fetchAudio(src);
-                    blob = new Blob([arrayBuffer as any]);
+                    const data = await Native.fetchAudio(src);
+                    blob = new Blob([new Uint8Array(data)]);
                 } else {
                     const res = await fetch(src);
                     if (!res.ok) throw new Error("Failed to download audio");
@@ -196,6 +264,7 @@ function TranscriptionModal(props: { modalProps: RenderModalProps, src: string,
                 const audioData = await decodeAudio(blob);
 
                 if (!active) return;
+                workerRef.current?.terminate();
                 workerRef.current = new TranscriptionWorker(
                     s => {
                         if (active) setStatus(s);
@@ -207,7 +276,10 @@ function TranscriptionModal(props: { modalProps: RenderModalProps, src: string,
                         }
                     },
                     err => {
-                        if (active) setError(String(err));
+                        if (active) {
+                            setError(String(err));
+                            setStatus("error");
+                        }
                     },
                     partial => {
                         if (active) setResult(partial);
@@ -223,7 +295,10 @@ function TranscriptionModal(props: { modalProps: RenderModalProps, src: string,
                     options.task
                 );
             } catch (err) {
-                if (active) setError(String(err));
+                if (active) {
+                    setError(String(err));
+                    setStatus("error");
+                }
             }
         })();
 
@@ -231,20 +306,16 @@ function TranscriptionModal(props: { modalProps: RenderModalProps, src: string,
             active = false;
             workerRef.current?.terminate();
         };
-    }, [src, retryCount]);
+    }, [retryCount]);
 
-    const retry = () => {
-        setError(null);
-        setStatus("initializing");
-        setResult(null);
-        setCopied(false);
-        setRetryCount(prev => prev + 1);
-    };
-
-    const formatTimestamp = (seconds: number) => {
-        const m = Math.floor(seconds / 60);
-        const s = Math.floor(seconds % 60);
-        return `${m.toString().padStart(2, "0")}:${s.toString().padStart(2, "0")}`;
+    const handleCopy = () => {
+        if (!result) return;
+        const text = showTimestamps
+            ? result.chunks.map(c => `[${formatTimestamp(c.timestamp[0])} - ${formatTimestamp(c.timestamp[1])}] ${c.text}`).join("\n")
+            : result.text;
+        copyToClipboard(text);
+        setCopied(true);
+        setTimeout(() => setCopied(false), 2000);
     };
 
     const displayText = result ? (
@@ -253,61 +324,46 @@ function TranscriptionModal(props: { modalProps: RenderModalProps, src: string,
             : result.text
     ) : "";
 
-    const actions: any[] = [];
-    if (error) {
-        actions.push({
-            text: "Retry",
-            variant: "primary",
-            onClick: retry
-        });
-    } else if (status === "complete" || (status === "transcribing" && result)) {
-        actions.push({
-            text: copied ? "Copied!" : "Copy to Clipboard",
-            variant: "primary",
-            disabled: status === "transcribing" || copied,
-            onClick: () => {
-                copyToClipboard(displayText);
-                setCopied(true);
-                setTimeout(() => setCopied(false), 2000);
-            }
-        });
-    }
-
     return (
         <Modal
             {...modalProps}
-            size="lg"
-            title="Transcription"
-            actions={actions}
+            size="md"
+            title={options.task === "translate" ? "Voice Message Translation" : "Voice Message Transcription"}
+            actions={[
+                ...(error ? [{
+                    text: "Retry",
+                    variant: "primary" as const,
+                    onClick: () => setRetryCount(c => c + 1)
+                }] : []),
+                ...(result ? [
+                    {
+                        text: showTimestamps ? "Hide Timestamps" : "Show Timestamps",
+                        variant: "secondary" as const,
+                        onClick: () => setShowTimestamps(!showTimestamps)
+                    },
+                    {
+                        text: copied ? "Copied!" : "Copy Text",
+                        variant: "secondary" as const,
+                        onClick: handleCopy
+                    }
+                ] : [])
+            ]}
         >
             <div className={cl("content")}>
                 {error ? (
-                    <Flex flexDirection="column" gap={16} style={{ padding: "20px 0" }}>
-                        <Heading tag="h3" style={{ color: "var(--red-360)" }}>Error</Heading>
-                        <Span style={{ whiteSpace: "pre-wrap" }}>{error}</Span>
-                    </Flex>
-                ) : (status === "complete" || (status === "transcribing" && result)) ? (
-                    <Flex flexDirection="column" gap={16} style={{ paddingBottom: "20px" }}>
-                        {status === "transcribing" && (
-                            <Span size="sm" color="text-muted">Transcribing in progress...</Span>
-                        )}
-                        <ScrollerAuto className={cl("result")}>
-                            <Span>{displayText}</Span>
-                        </ScrollerAuto>
-                        <Flex flexDirection="row" gap={12} alignItems="center">
-                            <div style={{ flexGrow: 1 }}>
-                                <FormSwitch
-                                    title="Show Timestamps"
-                                    value={showTimestamps}
-                                    onChange={setShowTimestamps}
-                                />
-                            </div>
-                        </Flex>
+                    <Flex flexDirection="column" alignItems="center" gap={12} style={{ padding: "32px 16px" }}>
+                        <Heading tag="h3" style={{ color: "var(--red-360)" }}>Transcription Failed</Heading>
+                        <BaseText size="sm" color="text-muted" style={{ textAlign: "center" }}>
+                            {error}
+                        </BaseText>
                     </Flex>
+                ) : displayText ? (
+                    <div className={cl("result")}>
+                        <BaseText size="md">{displayText}</BaseText>
+                    </div>
                 ) : (
-                    <Flex flexDirection="column" gap={16} style={{ padding: "20px 0", alignItems: "center" }}>
+                    <Flex flexDirection="column" alignItems="center" justifyContent="center" gap={16} style={{ height: "200px" }}>
                         <Heading tag="h3">
-                            {status === "initializing" && "Initializing..."}
                             {status === "downloading_audio" && "Downloading Audio..."}
                             {status === "processing_audio" && "Processing Audio..."}
                             {status === "loading" && "Loading Model..."}
@@ -320,9 +376,250 @@ function TranscriptionModal(props: { modalProps: RenderModalProps, src: string,
     );
 }
 
+function VoiceMessageTranscriber({ src }: { src: string; }) {
+    const { embed, maintainHorizontal, quantized, selectedModel } = settings.use(["embed", "maintainHorizontal", "quantized", "selectedModel"]);
+    const [isOpen, setIsOpen] = useState(false);
+    const [status, setStatus] = useState<string>("idle");
+    const [result, setResult] = useState<TranscriptionResult | null>(null);
+    const [showTimestamps, setShowTimestamps] = useState(false);
+    const [error, setError] = useState<string | null>(null);
+    const [copied, setCopied] = useState(false);
+
+    const workerRef = useRef<TranscriptionWorker | null>(null);
+    const activeRunId = useRef(0);
+    const buttonRef = useRef<HTMLButtonElement>(null);
+
+    useEffect(() => {
+        const parent = buttonRef.current?.parentElement;
+        if (!parent) return;
+        parent.classList.add("vc-transcription-container");
+        return () => {
+            parent.classList.remove("vc-transcription-container", "vc-transcription-has-embed", "vc-transcription-maintain-horizontal");
+        };
+    }, []);
+
+    useEffect(() => {
+        const parent = buttonRef.current?.parentElement;
+        if (!parent) return;
+        parent.classList.toggle("vc-transcription-has-embed", Boolean(embed && isOpen));
+        parent.classList.toggle("vc-transcription-maintain-horizontal", Boolean(embed && isOpen && maintainHorizontal));
+    }, [embed, isOpen, maintainHorizontal]);
+
+    useEffect(() => {
+        return () => {
+            activeRunId.current++;
+            workerRef.current?.terminate();
+            workerRef.current = null;
+        };
+    }, []);
+
+    const startTranscription = async () => {
+        const runId = ++activeRunId.current;
+        workerRef.current?.terminate();
+        workerRef.current = null;
+
+        setIsOpen(true);
+        setError(null);
+        setStatus("downloading_audio");
+
+        try {
+            let blob: Blob;
+            if (IS_DISCORD_DESKTOP || IS_EQUIBOP) {
+                const data = await Native.fetchAudio(src);
+                blob = new Blob([new Uint8Array(data)]);
+            } else {
+                const res = await fetch(src);
+                if (!res.ok) throw new Error("Failed to download audio");
+                blob = await res.blob();
+            }
+
+            if (runId !== activeRunId.current) return;
+
+            setStatus("processing_audio");
+            const audioData = await decodeAudio(blob);
+
+            if (runId !== activeRunId.current) return;
+
+            workerRef.current = new TranscriptionWorker(
+                s => {
+                    if (runId === activeRunId.current) setStatus(s);
+                },
+                out => {
+                    if (runId === activeRunId.current) {
+                        setResult(out);
+                        setStatus("complete");
+                    }
+                },
+                err => {
+                    if (runId === activeRunId.current) {
+                        setError(String(err));
+                        setStatus("error");
+                    }
+                },
+                partial => {
+                    if (runId === activeRunId.current) setResult(partial);
+                }
+            );
+
+            workerRef.current.run(
+                audioData,
+                selectedModel,
+                quantized,
+                undefined,
+                "transcribe"
+            );
+        } catch (err) {
+            if (runId === activeRunId.current) {
+                setError(String(err));
+                setStatus("error");
+            }
+        }
+    };
+
+    const handleClick = (e: React.MouseEvent) => {
+        e.stopPropagation();
+        if (embed) {
+            if (!isOpen) {
+                if (status === "idle" || status === "error") {
+                    startTranscription();
+                } else {
+                    setIsOpen(true);
+                }
+            } else {
+                setIsOpen(false);
+            }
+        } else {
+            openModal(modalProps => <LanguageSelectionModal modalProps={modalProps} src={src} />);
+        }
+    };
+
+    const displayText = result ? (
+        showTimestamps
+            ? result.chunks.map(c => `[${formatTimestamp(c.timestamp[0])} - ${formatTimestamp(c.timestamp[1])}] ${c.text}`).join("\n")
+            : result.text
+    ) : "";
+
+    const handleCopy = (e: React.MouseEvent) => {
+        e.stopPropagation();
+        if (!displayText) return;
+        copyToClipboard(displayText);
+        setCopied(true);
+        setTimeout(() => setCopied(false), 2000);
+    };
+
+    const isWorking = status === "downloading_audio" || status === "processing_audio" || status === "loading" || status === "transcribing";
+
+    return (
+        <>
+            <button
+                ref={buttonRef}
+                className={cl("button")}
+                style={{ backgroundColor: "transparent" }}
+                onClick={handleClick}
+                title="Transcribe Voice Message"
+            >
+                <ChannelListIcon colorClass={cl("icon")} />
+            </button>
+            {embed && isOpen && (
+                <div
+                    className={classes(
+                        cl("embed"),
+                        maintainHorizontal && cl("embed-maintain-horizontal")
+                    )}
+                    onClick={e => e.stopPropagation()}
+                >
+                    <div className={cl("embed-header")}>
+                        <div className={cl("embed-title")}>
+                            <span className={classes(cl("status-dot"), isWorking && cl("status-dot-active"), status === "error" && cl("status-dot-error"))} />
+                            <span>
+                                {status === "downloading_audio" && "Downloading Audio..."}
+                                {status === "processing_audio" && "Processing Audio..."}
+                                {status === "loading" && "Loading Model..."}
+                                {status === "transcribing" && "Transcribing..."}
+                                {status === "complete" && "Transcription"}
+                                {status === "error" && "Transcription Error"}
+                            </span>
+                        </div>
+                        <div className={cl("embed-actions")}>
+                            {result && (
+                                <>
+                                    <Tooltip text={showTimestamps ? "Hide Timestamps" : "Show Timestamps"}>
+                                        {props => (
+                                            <button
+                                                {...props}
+                                                className={classes(cl("action-btn"), showTimestamps && cl("action-btn-active"))}
+                                                onClick={e => {
+                                                    e.stopPropagation();
+                                                    setShowTimestamps(!showTimestamps);
+                                                }}
+                                            >
+                                                <ClockIcon width={14} height={14} />
+                                            </button>
+                                        )}
+                                    </Tooltip>
+                                    <Tooltip text={copied ? "Copied!" : "Copy Text"}>
+                                        {props => (
+                                            <button
+                                                {...props}
+                                                className={cl("action-btn")}
+                                                onClick={handleCopy}
+                                            >
+                                                {copied ? <CheckmarkIcon size={14} /> : <CopyIcon width={14} height={14} />}
+                                            </button>
+                                        )}
+                                    </Tooltip>
+                                </>
+                            )}
+                            <Tooltip text="Close">
+                                {props => (
+                                    <button
+                                        {...props}
+                                        className={cl("action-btn")}
+                                        onClick={e => {
+                                            e.stopPropagation();
+                                            setIsOpen(false);
+                                        }}
+                                    >
+                                        <CloseIcon size={14} />
+                                    </button>
+                                )}
+                            </Tooltip>
+                        </div>
+                    </div>
+                    <div className={cl("embed-body")}>
+                        {error ? (
+                            <Flex flexDirection="column" gap={8}>
+                                <Span size="xs" color="text-danger">{error}</Span>
+                                <Button
+                                    size="small"
+                                    variant="primary"
+                                    onClick={startTranscription}
+                                    style={{ alignSelf: "flex-start" }}
+                                >
+                                    Retry
+                                </Button>
+                            </Flex>
+                        ) : displayText ? (
+                            <ScrollerAuto className={cl("embed-text")}>
+                                <Span size="sm">{displayText}</Span>
+                            </ScrollerAuto>
+                        ) : (
+                            <Span size="xs" color="text-muted">
+                                {isWorking ? "Transcribing in progress..." : "Initializing..."}
+                            </Span>
+                        )}
+                    </div>
+                </div>
+            )}
+        </>
+    );
+}
+
+const VoiceMessageTranscriberWrapped = ErrorBoundary.wrap(VoiceMessageTranscriber, { noop: true });
+
 export default definePlugin({
     name: "VoiceMessageTranscriber",
-    authors: [Devs.TheSun],
+    authors: [Devs.TheSun, EquicordDevs.tt],
     description: "On-device transcriptions for voice messages powered by Whisper v3",
     tags: ["Chat", "Media", "Utility", "Voice"],
     patches: [
@@ -341,20 +638,12 @@ export default definePlugin({
             }
         },
     ],
-    set ManaBaseRadioGroup(value: any) {
+    set ManaBaseRadioGroup(value: typeof ManaBaseRadioGroup) {
         ManaBaseRadioGroup = value;
     },
     settings,
 
     button(src: string) {
-        return <button
-            className={cl("button")}
-            style={{ backgroundColor: "transparent" }}
-            onClick={e => {
-                e.stopPropagation();
-                openModal(modalProps => <LanguageSelectionModal modalProps={modalProps} src={src} />);
-            }}>
-            <ChannelListIcon colorClass={cl("icon")} />
-        </button>;
+        return <VoiceMessageTranscriberWrapped src={src} />;
     },
 });

src/equicordplugins/voiceMessageTranscriber.desktop/style.css

@@ -24,4 +24,175 @@
     user-select: text;
     max-height: 400px;
     overflow-y: auto;
+}
+
+.vc-transcription-model-option {
+    display: flex;
+    align-items: center;
+    gap: 8px;
+    width: 100%;
+}
+
+.vc-transcription-model-size {
+    font-size: 12px;
+    color: var(--text-muted);
+    opacity: 0.75;
+    flex-shrink: 0;
+}
+
+.vc-transcription-container:has(> .vc-transcription-embed),
+.vc-transcription-container.vc-transcription-has-embed {
+    flex-wrap: wrap !important;
+}
+
+.vc-transcription-container:has(> .vc-transcription-embed-maintain-horizontal),
+.vc-transcription-container.vc-transcription-maintain-horizontal {
+    flex-wrap: wrap !important;
+    width: fit-content !important;
+    max-width: 100% !important;
+}
+
+.vc-transcription-embed {
+    order: 9999 !important;
+    display: flex;
+    flex-direction: column;
+    width: 100%;
+    flex-basis: 100%;
+    min-width: 0;
+    margin-top: 8px;
+    border-radius: 8px;
+    background: rgb(0 0 0 / 25%);
+    border: 1px solid var(--border-subtle, rgb(255 255 255 / 8%));
+    backdrop-filter: blur(4px);
+    box-sizing: border-box;
+    overflow: hidden;
+    user-select: text;
+}
+
+.vc-transcription-embed-header {
+    display: flex;
+    justify-content: space-between;
+    align-items: center;
+    padding: 6px 10px;
+    background: rgb(0 0 0 / 15%);
+    border-bottom: 1px solid var(--border-subtle, rgb(255 255 255 / 6%));
+}
+
+.vc-transcription-embed-title {
+    display: flex;
+    align-items: center;
+    gap: 6px;
+    font-size: 11px;
+    font-weight: 600;
+    text-transform: uppercase;
+    letter-spacing: 0.04em;
+    color: var(--text-muted);
+}
+
+.vc-transcription-status-dot {
+    width: 6px;
+    height: 6px;
+    border-radius: 50%;
+    background-color: var(--text-muted);
+    transition: background-color 0.2s ease;
+}
+
+.vc-transcription-status-dot-active {
+    background-color: var(--main-color, var(--brand-experiment, var(--brand-500, #5865f2)));
+    animation: vc-transcription-pulse 1.5s infinite;
+}
+
+.vc-transcription-status-dot-error {
+    background-color: var(--text-feedback-critical, #ed4245);
+}
+
+@keyframes vc-transcription-pulse {
+    0%, 100% {
+        opacity: 1;
+        transform: scale(1);
+    }
+
+    50% {
+        opacity: 0.35;
+        transform: scale(0.85);
+    }
+}
+
+.vc-transcription-embed-actions {
+    display: flex;
+    align-items: center;
+    gap: 4px;
+}
+
+.vc-transcription-action-btn {
+    display: flex;
+    align-items: center;
+    justify-content: center;
+    width: 22px;
+    height: 22px;
+    padding: 0;
+    border: none;
+    border-radius: 4px;
+    background: transparent;
+    color: var(--interactive-icon-default, var(--text-muted));
+    cursor: pointer;
+    transition: color 0.15s ease, background-color 0.15s ease;
+}
+
+.vc-transcription-action-btn:hover {
+    color: var(--interactive-icon-hover, var(--text-normal));
+    background-color: var(--background-modifier-hover, rgb(255 255 255 / 8%));
+}
+
+.vc-transcription-action-btn-active {
+    color: var(--main-color, var(--brand-experiment, var(--brand-500, #5865f2)));
+    background-color: var(--background-modifier-selected, rgb(255 255 255 / 12%));
+}
+
+.vc-transcription-embed-body {
+    padding: 8px 12px;
+    max-height: 180px;
+    overflow-y: auto;
+}
+
+.vc-transcription-embed-text {
+    white-space: pre-wrap;
+    user-select: text;
+}
+
+.vc-transcription-embed-maintain-horizontal {
+    order: 9999 !important;
+    width: 0 !important;
+    min-width: 100% !important;
+    max-width: 100% !important;
+    flex-basis: 100% !important;
+    flex-grow: 1 !important;
+    box-sizing: border-box !important;
+    margin-left: 0 !important;
+    margin-right: 0 !important;
+}
+
+.vc-transcription-embed-maintain-horizontal .vc-transcription-embed-header {
+    width: 100% !important;
+    min-width: 0 !important;
+    max-width: 100% !important;
+    box-sizing: border-box !important;
+}
+
+.vc-transcription-embed-maintain-horizontal .vc-transcription-embed-body {
+    width: 100% !important;
+    min-width: 0 !important;
+    max-width: 100% !important;
+    max-height: 280px;
+    box-sizing: border-box !important;
+    overflow-wrap: anywhere;
+}
+
+.vc-transcription-embed-maintain-horizontal .vc-transcription-embed-text {
+    width: 100% !important;
+    min-width: 0 !important;
+    max-width: 100% !important;
+    box-sizing: border-box !important;
+    overflow-wrap: anywhere;
+    display: block;
 }

src/equicordplugins/voiceMessageTranscriber.desktop/utils.ts

@@ -112,7 +112,11 @@ export const LANGUAGES = {
 
 export const cl = classNameFactory("vc-transcription-");
 
-const getAudioContext = () => new (window.AudioContext || (window as any).webkitAudioContext)({ sampleRate: 16000 });
+const AudioContextClass = window.AudioContext || (window as Window & { webkitAudioContext?: typeof AudioContext }).webkitAudioContext;
+const getAudioContext = () => {
+    if (!AudioContextClass) throw new Error("AudioContext is not supported in this environment");
+    return new AudioContextClass({ sampleRate: 16000 });
+};
 export async function decodeAudio(blob: Blob): Promise<Float32Array> {
     const arrayBuffer = await blob.arrayBuffer();
     const audioContext = getAudioContext();
@@ -258,18 +262,28 @@ async function runTranscription({ audio, model, quantized, language, task }) {
 }
 `;
 
+export interface TranscriptionChunk {
+    timestamp: [number, number];
+    text: string;
+}
+
+export interface TranscriptionResult {
+    text: string;
+    chunks: TranscriptionChunk[];
+}
+
 export class TranscriptionWorker {
     private worker: Worker;
     private onStatus: (status: string) => void;
-    private onComplete: (output: any) => void;
-    private onError: (error: any) => void;
-    private onPartial: (output: any) => void;
+    private onComplete: (output: TranscriptionResult) => void;
+    private onError: (error: unknown) => void;
+    private onPartial: (output: TranscriptionResult) => void;
 
     constructor(
         onStatus: (status: string) => void,
-        onComplete: (output: any) => void,
-        onError: (error: any) => void,
-        onPartial: (output: any) => void
+        onComplete: (output: TranscriptionResult) => void,
+        onError: (error: unknown) => void,
+        onPartial: (output: TranscriptionResult) => void
     ) {
         this.onStatus = onStatus;
         this.onComplete = onComplete;