VoiceStudio is the open-source, fully-local ElevenLabs alternative — voice cloning, voice design, video dubbing, dictation, transcription & audiobook creation in 646 languages.
Python aiaudiobookcudadubbingelevenlabs-alternativehuggingfacelocal-firstmlxomnivoice-studiospeech-to-texttext-to-speechtranscriptiontranslatettsvoice-aivoice-cloningvoice-generationvoicestudioworkflow
/**
* Settings → Capture → Voice panel.
*
* The home of the live-dictation controls (the "Voice" card in the screenshot):
*
* 1. Enable Voice Dictation — master toggle. Subtitle shows the REAL
* effective dictation shortcut (same live source as HotkeyTab); changing it still
* lives in HotkeyTab below.
* 2. Dictation Mode — Toggle / Hold segmented control. Toggle = press once to
* start, again to stop. Hold = dictate while the key is held.
* 3. Speech Model — a dropdown of the seven sherpa-onnx dictation models. The
* collapsed control shows the selected model's name + description; expanded,
* each row shows offline/streaming + recommended badges, the size, a
* one-line description, a checkmark on the selected one, and a trash icon to
* delete an installed model. Not-installed models show a Download action and
* inline download progress while installing.
*
* Prefs are the backend `dictation.*` namespace (GET/POST /dictation/prefs),
* mirrored into the zustand store; model install/delete/progress REUSE the
* model-store endpoints + SSE stream the ModelStoreTab already drives, so a
* model installed here shows installed there and vice-versa.
*
* Cross-platform: every control behaves identically on macOS / Windows / Linux.
* The shortcut read no-ops gracefully in the web UI (no Tauri) and falls back to
* the documented Ctrl/Cmd+Shift+Space default label.
*/
import React, { useEffect, useMemo, useRef, useState } from 'react';
import { Mic, Check, Download, Trash2, ChevronDown, Loader, AlertTriangle } from 'lucide-react';
import { useTranslation } from 'react-i18next';
import { toast } from 'react-hot-toast';
import { useAppStore } from '../../store';
import { apiJson } from '../../api/client';
import { useInstallModel, useDeleteModel } from '../../api/hooks';
import { setupDownloadStreamUrl } from '../../api/setup';
import { isTauri as _isTauri } from '../../utils/media';
import { Badge, Progress, Segmented } from '../../ui';
import { useEffectiveDictationShortcut } from '../../hooks/useEffectiveDictationShortcut';
import { SettingsSection, SettingRow, SettingsToggle } from './primitives';
/** Native confirm dialog in Tauri, window.confirm in the web UI. Mirrors the
* `askConfirm` helper in Settings.jsx (kept local to avoid coupling). */
async function askConfirm(message, title = 'Confirm') {
if (_isTauri) {
try {
const { ask } = await import('@tauri-apps/plugin-dialog');
return ask(message, { title, kind: 'warning' });
} catch {
/* dialog plugin unavailable — fall through */
}
}
return Promise.resolve(typeof window !== 'undefined' ? window.confirm(message) : true);
}
/** Format a model's download size for display (e.g. "180 MB", "1.2 GB"). */
function fmtSize(sizeGb) {
if (sizeGb == null) return '';
if (sizeGb < 1) return `${Math.round(sizeGb * 1000)} MB`;
return `${sizeGb.toFixed(sizeGb < 10 ? 1 : 0)} GB`;
}
export default function VoicePanel() {
const { t } = useTranslation();
const enabled = useAppStore((s) => s.dictationEnabled);
const setEnabled = useAppStore((s) => s.setDictationEnabled);
const mode = useAppStore((s) => s.dictationMode);
const setMode = useAppStore((s) => s.setDictationMode);
const modelId = useAppStore((s) => s.dictationModelId);
const setModelId = useAppStore((s) => s.setDictationModelId);
const loadPrefs = useAppStore((s) => s.loadDictationPrefs);
const [models, setModels] = useState([]);
const [engineAvailable, setEngineAvailable] = useState(true);
const [engineReason, setEngineReason] = useState(null);
const [loading, setLoading] = useState(true);
const [open, setOpen] = useState(false);
const { info: shortcut } = useEffectiveDictationShortcut(_isTauri);
// Per-repo download runtime, keyed by repo_id, driven by the shared SSE
// progress stream (same events ModelStoreTab consumes):
// { [repo_id]: { phase, pct } }
const [rowState, setRowState] = useState({});
const esRef = useRef(null);
const dropdownRef = useRef(null);
const installMutation = useInstallModel();
const deleteMutation = useDeleteModel();
// Hydrate prefs from the backend on mount (write-through setters keep them in
// sync after that).
useEffect(() => {
loadPrefs();
}, [loadPrefs]);
const loadModels = React.useCallback(async () => {
try {
const data = await apiJson('/dictation/models');
setModels(Array.isArray(data?.models) ? data.models : []);
setEngineAvailable(data?.engine_available !== false);
setEngineReason(data?.engine_reason || null);
} catch {
setModels([]);
} finally {
setLoading(false);
}
}, []);
useEffect(() => {
loadModels();
}, [loadModels]);
// Subscribe to the model-store SSE so install progress shows inline here too.
// We only track the aggregate percent + lifecycle phase per repo — the full
// per-file accounting lives in ModelStoreTab; here a single bar is enough.
useEffect(() => {
const es = new EventSource(setupDownloadStreamUrl());
esRef.current = es;
es.onmessage = (evt) => {
try {
const ev = JSON.parse(evt.data);
if (!ev?.repo_id) return;
setRowState((prev) => {
const cur = prev[ev.repo_id] || {};
if (ev.phase === 'install_start')
return { ...prev, [ev.repo_id]: { phase: 'active', pct: 0 } };
if (ev.phase === 'install_done')
return { ...prev, [ev.repo_id]: { phase: 'install_done', pct: 100 } };
if (ev.phase === 'install_error')
return { ...prev, [ev.repo_id]: { phase: 'install_error', error: ev.error } };
if (ev.phase === 'install_cancelled')
return { ...prev, [ev.repo_id]: { phase: 'install_cancelled' } };
if (ev.phase === 'delete_done')
return { ...prev, [ev.repo_id]: { phase: 'delete_done' } };
if (ev.phase === 'aggregate') {
const total = ev.total_bytes || 0;
const bytePct = total > 0 ? (ev.bytes_done / total) * 100 : 0;
const ft = ev.files_total || 0;
const filePct = ft > 0 ? ((ev.files_done || 0) / ft) * 100 : 0;
return { ...prev, [ev.repo_id]: { phase: 'active', pct: Math.max(bytePct, filePct) } };
}
// Plain per-file tqdm event: approximate with its pct if no aggregate yet.
if (ev.pct != null && cur.phase !== 'active') {
return { ...prev, [ev.repo_id]: { phase: 'active', pct: ev.pct } };
}
return prev;
});
} catch {
/* keepalive */
}
};
return () => es.close();
}, []);
// When an install/delete terminates, refresh the model list so the installed
// flag flips, then clear the terminal row so it reverts to the list state.
useEffect(() => {
const term = Object.entries(rowState).find(([, s]) =>
['install_done', 'delete_done', 'install_error', 'install_cancelled'].includes(s.phase),
);
if (!term) return;
const id = setTimeout(() => {
loadModels();
setRowState((prev) => {
const n = { ...prev };
delete n[term[0]];
return n;
});
}, 800);
return () => clearTimeout(id);
}, [rowState, loadModels]);
// Close the dropdown on outside-click / Escape.
useEffect(() => {
if (!open) return;
const onDown = (e) => {
if (dropdownRef.current && !dropdownRef.current.contains(e.target)) setOpen(false);
};
const onKey = (e) => {
if (e.key === 'Escape') setOpen(false);
};
window.addEventListener('mousedown', onDown);
window.addEventListener('keydown', onKey);
return () => {
window.removeEventListener('mousedown', onDown);
window.removeEventListener('keydown', onKey);
};
}, [open]);
const selected = useMemo(
() => models.find((m) => m.id === modelId) || models.find((m) => m.recommended) || models[0],
[models, modelId],
);
const onInstall = async (repoId) => {
setRowState((p) => ({ ...p, [repoId]: { phase: 'active', pct: 0 } }));
try {
await installMutation.mutateAsync(repoId);
} catch (e) {
toast.error(e?.message || String(e));
setRowState((p) => {
const n = { ...p };
delete n[repoId];
return n;
});
}
};
const onDelete = async (model) => {
const ok = await askConfirm(
t('voicePanel.delete_confirm', { label: model.label }),
t('voicePanel.delete_confirm_title'),
);
if (!ok) return;
setRowState((p) => ({ ...p, [model.repo_id]: { phase: 'deleting' } }));
try {
await deleteMutation.mutateAsync(model.repo_id);
} catch (e) {
toast.error(e?.message || String(e));
setRowState((p) => {
const n = { ...p };
delete n[model.repo_id];
return n;
});
}
};
// Picking a model: write the pref. If it isn't installed yet, also kick off
// the download — the user gets the model the moment it lands, and the live
// socket will pick it up via the persisted pref.
const onPick = (model) => {
setModelId(model.id);
setOpen(false);
if (!model.installed && !rowState[model.repo_id]) {
onInstall(model.repo_id);
}
};
// Human-readable, i18n-keyed description per model id (one line each). Falls
// back to the backend `languages` string when a key is missing.
const modelDesc = (m) => t(`voicePanel.model_desc.${m.id}`, { defaultValue: m.languages || '' });
const shortcutLabel = shortcut.display;
return (
<SettingsSection
className="voicepanel"
icon={Mic}
accent="var(--chrome-accent)"
title={t('voicePanel.title')}
description={t('voicePanel.subtitle')}
>
{!engineAvailable && (
<div
className="mb-[var(--space-4)] flex items-center gap-[var(--space-3)] rounded-[var(--chrome-radius-pill)] [border:1px_solid_color-mix(in_srgb,var(--chrome-severity-warn)_32%,transparent)] bg-[color-mix(in_srgb,var(--chrome-severity-warn)_12%,transparent)] px-[var(--space-4)] py-[var(--space-3)] text-[length:var(--text-sm)] text-[var(--chrome-severity-warn)]"
role="alert"
>
<AlertTriangle size={13} />
<span>{engineReason || t('voicePanel.engine_unavailable')}</span>
</div>
)}
{/* 1 — Enable Voice Dictation */}
<SettingRow
title={t('voicePanel.enable_label')}
subtitle={t('voicePanel.enable_sub', { shortcut: shortcutLabel })}
control={
<SettingsToggle
checked={enabled}
onChange={setEnabled}
aria-label={t('voicePanel.enable_label')}
/>
}
/>
{/* 2 — Dictation Mode */}
<SettingRow
title={t('voicePanel.mode_label')}
subtitle={t('voicePanel.mode_sub')}
control={
<Segmented
size="sm"
value={mode}
onChange={(v) => setMode(v)}
items={[
{ value: 'toggle', label: t('voicePanel.mode_toggle') },
{ value: 'hold', label: t('voicePanel.mode_hold') },
]}
/>
}
/>
{/* 3 — Speech Model */}
<SettingRow
className="voicepanel__row--model"
align="start"
title={t('voicePanel.model_label')}
subtitle={selected ? modelDesc(selected) : t('voicePanel.model_sub')}
control={
<div
className="relative w-auto min-w-[200px] max-w-[280px] flex-none text-left"
ref={dropdownRef}
>
<button
type="button"
className="flex w-full cursor-pointer items-center justify-between gap-[8px] rounded-[var(--chrome-radius-pill)] [border:1px_solid_var(--chrome-border,rgba(255,255,255,0.1))] bg-[var(--chrome-input-bg,rgba(255,255,255,0.04))] px-[var(--space-4)] py-[var(--space-3)] text-left text-[length:var(--text-md)] text-[var(--chrome-fg)] disabled:cursor-not-allowed disabled:opacity-60"
aria-haspopup="listbox"
aria-expanded={open}
disabled={loading || models.length === 0}
onClick={() => setOpen((o) => !o)}
data-testid="dictation-model-trigger"
>
<span className="truncate">{selected ? selected.label : t('common.loading')}</span>
<ChevronDown size={14} className={`voicepanel__dd-chev ${open ? 'is-open' : ''}`} />
</button>
{open && (
<ul
className="absolute right-0 top-[calc(100%_+_6px)] z-40 m-0 max-h-[380px] w-[360px] max-w-[80vw] list-none overflow-y-auto rounded-[10px] [border:1px_solid_var(--chrome-border,rgba(255,255,255,0.1))] bg-[var(--chrome-menu-bg,#1d1d22)] p-[5px] shadow-[0_12px_32px_rgba(0,0,0,0.45)]"
role="listbox"
aria-label={t('voicePanel.model_label')}
>
{models.map((m) => {
const rs = rowState[m.repo_id];
const installing = rs && rs.phase === 'active';
const deleting = rs && rs.phase === 'deleting';
const isSel = m.id === selected?.id;
return (
<li
key={m.id}
className={`flex items-start gap-[4px] rounded-[8px] hover:bg-[var(--chrome-hover-bg,rgba(255,255,255,0.08))] ${
isSel ? 'bg-[var(--chrome-hover-bg,rgba(255,255,255,0.06))]' : ''
}`}
role="option"
aria-selected={isSel}
>
<button
type="button"
className="flex min-w-0 flex-[1_1_auto] cursor-pointer items-start gap-[8px] border-0 bg-transparent pb-[9px] pl-[8px] pr-[4px] pt-[9px] text-left text-[var(--chrome-fg)]"
onClick={() => onPick(m)}
data-testid={`dictation-model-${m.id}`}
>
<span className="mt-[1px] w-[16px] flex-none text-[var(--chrome-accent,#83a598)]">
{isSel && <Check size={14} />}
</span>
<span className="flex min-w-0 flex-col gap-[4px]">
<span className="flex flex-wrap items-center gap-[6px]">
<span className="text-[0.86rem] font-semibold">{m.label}</span>
<Badge tone="neutral" size="xs">
{m.tag === 'streaming'
? t('voicePanel.badge_streaming')
: t('voicePanel.badge_offline')}
</Badge>
{m.recommended && (
<Badge tone="success" size="xs">
{t('voicePanel.badge_recommended')}
</Badge>
)}
<span className="ml-auto text-[0.74rem] text-[var(--chrome-fg-muted)]">
{fmtSize(m.size_gb)}
</span>
</span>
<span className="text-[0.76rem] leading-[1.4] text-[var(--chrome-fg-muted)]">
{modelDesc(m)}
</span>
{installing && (
<span className="voicepanel__dd-progress">
<Progress value={rs.pct ?? null} tone="brand" size="xs" />
<span className="flex-none text-[0.72rem] text-[var(--chrome-fg-muted)]">
{rs.pct != null && rs.pct > 0
? t('voicePanel.downloading_pct', { pct: Math.round(rs.pct) })
: t('voicePanel.downloading')}
</span>
</span>
)}
</span>
</button>
{/* Right-edge affordance: download (not installed), spinner
(installing/deleting), or delete (installed). */}
<span className="flex flex-none items-center pb-0 pl-0 pr-[6px] pt-[9px]">
{installing || deleting ? (
<Loader size={14} className="voicepanel__spin" />
) : m.installed ? (
<button
type="button"
className="inline-flex h-[26px] w-[26px] cursor-pointer items-center justify-center rounded-[6px] border-0 bg-transparent p-0 text-[var(--chrome-fg-muted)] hover:bg-[var(--chrome-hover-bg,rgba(255,255,255,0.1))] hover:text-[var(--chrome-fg)]"
title={t('voicePanel.delete_model')}
aria-label={t('voicePanel.delete_model')}
onClick={(e) => {
e.stopPropagation();
onDelete(m);
}}
data-testid={`dictation-delete-${m.id}`}
>
<Trash2 size={13} />
</button>
) : (
<button
type="button"
className="inline-flex h-[26px] w-[26px] cursor-pointer items-center justify-center rounded-[6px] border-0 bg-transparent p-0 text-[var(--chrome-fg-muted)] hover:bg-[var(--chrome-hover-bg,rgba(255,255,255,0.1))] hover:text-[var(--chrome-fg)]"
title={t('voicePanel.download_model')}
aria-label={t('voicePanel.download_model')}
onClick={(e) => {
e.stopPropagation();
onInstall(m.repo_id);
}}
data-testid={`dictation-install-${m.id}`}
>
<Download size={13} />
</button>
)}
</span>
</li>
);
})}
</ul>
)}
</div>
}
/>
</SettingsSection>
);
}