feat(capture): hold mobile + button to record voice thoughts directly to inbox

Doku/ui-patterns.md

@@ -68,3 +68,14 @@ Komponente: [CountBadge.tsx](../frontend/src/components/CountBadge.tsx), CSS-Kla
 - `index.html` enthält in `#root` einen reinen HTML-Platzhalter (vollflächig Blau, Logo in der Mitte, kein Text). React ersetzt ihn beim ersten Rendern, dazwischen blitzt nichts Weißes auf.
 - `meta theme-color` startet blau und wird von [`theme.ts`](../frontend/src/lib/theme.ts) nach dem Start und bei jedem Theme-Wechsel auf `--bg-app` des aktiven Themes gesetzt, damit die Statusleiste zum Header passt.
 
+## +-Button: Tippen und Halten
+
+Komponente: [HoldToRecordFab.tsx](../frontend/src/features/capture/HoldToRecordFab.tsx)
+
+- **Kurz tippen (< 350 ms):** Führt wie bisher die primäre Aktion der aktuellen Ansicht (`primaryAction`) oder den mobilen Erfassungsdialog (`openMobileTriageCapture`) aus.
+- **Gedrückt halten (≥ 350 ms):** Startet direkt eine Sprachaufnahme über die Web Speech API (`de-DE`, `continuous`, `interimResults`). Der Button wird zum pulsierenden Mikrofon (`bg-status-err`), darüber zeigt eine schwebende Pille „Ich höre zu… (loslassen speichert)“ samt Live-Transkript.
+- **Loslassen:** Stoppt die Aufnahme, wartet kurz auf das letzte Wort und speichert den Text unabhängig von der aktiven Ansicht immer als Gedanken in die Inbox (`source: 'voice'`, `target: 'inbox'`).
+- **Nach oben ziehen (> 60 px):** Schaltet die Pille in den Abbruch-Zustand („🗑️ Zum Abbrechen loslassen“); Loslassen dort verwirft die Aufnahme (`abort()`).
+- **Fallback:** Unterstützt der Browser keine Spracherkennung oder ist das Mikrofon gesperrt (`not-allowed`), öffnet Halten wie beim Tippen den Dialog und blendet eine kurze Statusmeldung über der Bottom-Leiste ein.
+
+

frontend/src/App.tsx

@@ -44,6 +44,7 @@ import {
   ThemeId,
 } from './lib/theme';
 import { InstantCaptureBar } from './features/capture/InstantCaptureBar';
+import { HoldToRecordFab } from './features/capture/HoldToRecordFab';
 import { CountBadge } from './components/CountBadge';
 import { TriageDeck } from './features/triage/TriageDeck';
 import { TasksView } from './features/triage/TasksView';
@@ -1174,9 +1175,8 @@ export const App: React.FC = () => {
 
         {/* Col 3 (EXACT CENTER): + Quick-Add FAB */}
         <div className="flex items-center justify-center">
-          <button
+          <HoldToRecordFab
-            type="button"
+            onTap={() => {
-            onClick={() => {
               triggerHaptic(15);
               if (primaryAction) {
                 primaryAction();
@@ -1184,11 +1184,11 @@ export const App: React.FC = () => {
                 openMobileTriageCapture();
               }
             }}
-            className="w-14 h-14 -mt-6 rounded-full bg-accent text-on-accent shadow-lg flex items-center justify-center active:scale-90 transition-transform border-4 border-panel cursor-pointer"
+            onTranscript={async (text) => {
-            title="Schnell erfassen (+)"
+              const thought = await api.captureThought(text, 'normal', 'voice', '', { target: 'inbox' });
-          >
+              handleCaptured(thought, 'inbox');
-            <Plus className="w-7 h-7 stroke-[2.5]" />
+            }}
-          </button>
+          />
         </div>
 
         {/* Col 4: Aufgaben */}

frontend/src/features/capture/HoldToRecordFab.tsx

@@ -0,0 +1,359 @@
+import React, { useEffect, useRef, useState } from 'react';
+import { Mic, Plus } from 'lucide-react';
+import { triggerHaptic } from '../../lib/api';
+
+export interface HoldToRecordFabProps {
+  onTap: () => void;
+  onTranscript: (text: string) => Promise<void> | void;
+}
+
+const HOLD_DELAY_MS = 350;
+const CANCEL_DRAG_PX = 60;
+const STOP_TIMEOUT_MS = 1500;
+const TOAST_DURATION_MS = 2500;
+
+export const HoldToRecordFab: React.FC<HoldToRecordFabProps> = ({ onTap, onTranscript }) => {
+  const [isRecording, setIsRecording] = useState(false);
+  const [isCancelZone, setIsCancelZone] = useState(false);
+  const [liveTranscript, setLiveTranscript] = useState('');
+  const [toastMessage, setToastMessage] = useState<string | null>(null);
+
+  const isPressedRef = useRef(false);
+  const didEnterHoldRef = useRef(false);
+  const fallbackHandledRef = useRef(false);
+  const isCancelZoneRef = useRef(false);
+  const abortedRef = useRef(false);
+  const isStoppingRef = useRef(false);
+  const hasEndedRef = useRef(false);
+  const finalizedRef = useRef(false);
+  const startYRef = useRef(0);
+  const transcriptRef = useRef('');
+
+  const recognitionRef = useRef<any>(null);
+  const holdTimerRef = useRef<number | null>(null);
+  const stopTimerRef = useRef<number | null>(null);
+  const toastTimerRef = useRef<number | null>(null);
+
+  const clearHoldTimer = () => {
+    if (holdTimerRef.current !== null) {
+      window.clearTimeout(holdTimerRef.current);
+      holdTimerRef.current = null;
+    }
+  };
+
+  const clearStopTimer = () => {
+    if (stopTimerRef.current !== null) {
+      window.clearTimeout(stopTimerRef.current);
+      stopTimerRef.current = null;
+    }
+  };
+
+  const showToast = (msg: string) => {
+    if (toastTimerRef.current !== null) {
+      window.clearTimeout(toastTimerRef.current);
+    }
+    setToastMessage(msg);
+    toastTimerRef.current = window.setTimeout(() => {
+      setToastMessage(null);
+      toastTimerRef.current = null;
+    }, TOAST_DURATION_MS);
+  };
+
+  useEffect(() => {
+    return () => {
+      clearHoldTimer();
+      clearStopTimer();
+      if (toastTimerRef.current !== null) {
+        window.clearTimeout(toastTimerRef.current);
+      }
+      if (recognitionRef.current) {
+        try {
+          recognitionRef.current.abort?.();
+        } catch {
+          // ignore cleanup errors
+        }
+      }
+    };
+  }, []);
+
+  const finalizeRecording = async () => {
+    if (finalizedRef.current) return;
+    finalizedRef.current = true;
+    clearStopTimer();
+    recognitionRef.current = null;
+
+    if (abortedRef.current) return;
+
+    const text = transcriptRef.current.trim();
+    setLiveTranscript('');
+    if (!text) {
+      showToast('Nichts erkannt');
+      return;
+    }
+
+    triggerHaptic(25);
+    showToast('Im Quick Dump gespeichert');
+    try {
+      await onTranscript(text);
+    } catch {
+      showToast('Fehler beim Speichern');
+    }
+  };
+
+  const startVoiceRecording = () => {
+    const SpeechRecognition =
+      (window as any).SpeechRecognition || (window as any).webkitSpeechRecognition;
+
+    if (!SpeechRecognition) {
+      fallbackHandledRef.current = true;
+      showToast('Spracheingabe wird von diesem Browser nicht unterstützt');
+      onTap();
+      return;
+    }
+
+    try {
+      const recognition = new SpeechRecognition();
+      recognition.lang = 'de-DE';
+      recognition.interimResults = true;
+      recognition.continuous = true;
+
+      transcriptRef.current = '';
+      abortedRef.current = false;
+      isStoppingRef.current = false;
+      hasEndedRef.current = false;
+      finalizedRef.current = false;
+      isCancelZoneRef.current = false;
+
+      recognition.onresult = (event: any) => {
+        let combined = '';
+        for (let i = 0; i < event.results.length; i++) {
+          combined += event.results[i][0].transcript;
+        }
+        transcriptRef.current = combined;
+        setLiveTranscript(combined);
+      };
+
+      recognition.onerror = (event: any) => {
+        const errCode = event?.error;
+        if (errCode === 'not-allowed' || errCode === 'service-not-allowed') {
+          abortedRef.current = true;
+          clearStopTimer();
+          setIsRecording(false);
+          setIsCancelZone(false);
+          setLiveTranscript('');
+          if (!fallbackHandledRef.current) {
+            fallbackHandledRef.current = true;
+            showToast('Mikrofon nicht freigegeben');
+            onTap();
+          }
+          return;
+        }
+        if (isStoppingRef.current) {
+          void finalizeRecording();
+        }
+      };
+
+      recognition.onend = () => {
+        hasEndedRef.current = true;
+        if (isStoppingRef.current) {
+          void finalizeRecording();
+        }
+      };
+
+      recognitionRef.current = recognition;
+      setLiveTranscript('');
+      setIsCancelZone(false);
+      setIsRecording(true);
+      triggerHaptic([15, 60, 25]);
+      recognition.start();
+    } catch {
+      fallbackHandledRef.current = true;
+      setIsRecording(false);
+      showToast('Spracheingabe wird von diesem Browser nicht unterstützt');
+      onTap();
+    }
+  };
+
+  const handlePointerDown = (e: React.PointerEvent<HTMLButtonElement>) => {
+    if (e.pointerType === 'mouse' && e.button !== 0) return;
+
+    try {
+      e.currentTarget.setPointerCapture?.(e.pointerId);
+    } catch {
+      // ignore if unsupported in environment
+    }
+
+    clearHoldTimer();
+    clearStopTimer();
+
+    isPressedRef.current = true;
+    didEnterHoldRef.current = false;
+    fallbackHandledRef.current = false;
+    isCancelZoneRef.current = false;
+    startYRef.current = e.clientY;
+
+    holdTimerRef.current = window.setTimeout(() => {
+      holdTimerRef.current = null;
+      if (!isPressedRef.current) return;
+      didEnterHoldRef.current = true;
+      startVoiceRecording();
+    }, HOLD_DELAY_MS);
+  };
+
+  const handlePointerMove = (e: React.PointerEvent<HTMLButtonElement>) => {
+    if (!isPressedRef.current || !didEnterHoldRef.current || fallbackHandledRef.current) return;
+    const deltaUp = startYRef.current - e.clientY;
+    const inCancel = deltaUp > CANCEL_DRAG_PX;
+    if (inCancel !== isCancelZoneRef.current) {
+      isCancelZoneRef.current = inCancel;
+      setIsCancelZone(inCancel);
+      triggerHaptic(10);
+    }
+  };
+
+  const handlePointerUp = (e: React.PointerEvent<HTMLButtonElement>) => {
+    try {
+      e.currentTarget.releasePointerCapture?.(e.pointerId);
+    } catch {
+      // ignore
+    }
+
+    if (!isPressedRef.current) return;
+    isPressedRef.current = false;
+    clearHoldTimer();
+
+    if (!didEnterHoldRef.current) {
+      onTap();
+      return;
+    }
+
+    if (fallbackHandledRef.current) {
+      return;
+    }
+
+    setIsRecording(false);
+    setIsCancelZone(false);
+
+    if (isCancelZoneRef.current) {
+      abortedRef.current = true;
+      clearStopTimer();
+      setLiveTranscript('');
+      triggerHaptic(15);
+      try {
+        recognitionRef.current?.abort?.();
+      } catch {
+        // ignore
+      }
+      recognitionRef.current = null;
+      return;
+    }
+
+    isStoppingRef.current = true;
+    if (hasEndedRef.current || !recognitionRef.current) {
+      void finalizeRecording();
+      return;
+    }
+
+    stopTimerRef.current = window.setTimeout(() => {
+      void finalizeRecording();
+    }, STOP_TIMEOUT_MS);
+
+    try {
+      recognitionRef.current.stop();
+    } catch {
+      void finalizeRecording();
+    }
+  };
+
+  const handlePointerCancel = (e: React.PointerEvent<HTMLButtonElement>) => {
+    try {
+      e.currentTarget.releasePointerCapture?.(e.pointerId);
+    } catch {
+      // ignore
+    }
+
+    if (!isPressedRef.current) return;
+    isPressedRef.current = false;
+    clearHoldTimer();
+    clearStopTimer();
+
+    if (didEnterHoldRef.current && !fallbackHandledRef.current) {
+      abortedRef.current = true;
+      setIsRecording(false);
+      setIsCancelZone(false);
+      setLiveTranscript('');
+      try {
+        recognitionRef.current?.abort?.();
+      } catch {
+        // ignore
+      }
+      recognitionRef.current = null;
+    }
+  };
+
+  const displayedTranscript =
+    liveTranscript.length > 180 ? `…${liveTranscript.slice(-180)}` : liveTranscript;
+
+  return (
+    <>
+      {isRecording && (
+        <div
+          className={
+            isCancelZone
+              ? 'fixed left-1/2 -translate-x-1/2 bottom-[calc(5.5rem+env(safe-area-inset-bottom))] z-50 w-max max-w-[calc(100vw-1.5rem)] px-4 py-2.5 rounded-card bg-panel-elevated border border-status-err-border text-status-err shadow-panel pointer-events-none text-center'
+              : 'fixed left-1/2 -translate-x-1/2 bottom-[calc(5.5rem+env(safe-area-inset-bottom))] z-50 w-max max-w-[calc(100vw-1.5rem)] px-4 py-2.5 rounded-card bg-panel-elevated border border-border text-main shadow-panel pointer-events-none text-center'
+          }
+        >
+          {isCancelZone ? (
+            <div className="text-xs font-semibold">🗑️ Zum Abbrechen loslassen</div>
+          ) : (
+            <div className="space-y-1">
+              <div className="text-xs font-semibold text-main">
+                Ich höre zu… (loslassen speichert)
+              </div>
+              {displayedTranscript ? (
+                <div className="text-xs text-muted break-words max-w-[calc(100vw-3rem)]">
+                  {displayedTranscript}
+                </div>
+              ) : (
+                <div className="text-[11px] text-faint">Nach oben ziehen zum Abbrechen</div>
+              )}
+            </div>
+          )}
+        </div>
+      )}
+
+      {!isRecording && toastMessage && (
+        <div
+          role="status"
+          aria-live="polite"
+          className="fixed left-1/2 -translate-x-1/2 bottom-[calc(5.5rem+env(safe-area-inset-bottom))] z-50 w-max max-w-[calc(100vw-1.5rem)] px-4 py-2 rounded-pill bg-panel-elevated border border-border text-main text-xs font-medium shadow-panel pointer-events-none text-center"
+        >
+          {toastMessage}
+        </div>
+      )}
+
+      <button
+        type="button"
+        onPointerDown={handlePointerDown}
+        onPointerMove={handlePointerMove}
+        onPointerUp={handlePointerUp}
+        onPointerCancel={handlePointerCancel}
+        onContextMenu={(e) => e.preventDefault()}
+        className={
+          isRecording
+            ? 'w-14 h-14 -mt-6 rounded-full bg-status-err text-on-accent shadow-lg flex items-center justify-center transition-transform border-4 border-panel cursor-pointer touch-none select-none animate-pulse'
+            : 'w-14 h-14 -mt-6 rounded-full bg-accent text-on-accent shadow-lg flex items-center justify-center active:scale-90 transition-transform border-4 border-panel cursor-pointer touch-none select-none'
+        }
+        title="Schnell erfassen (+)"
+        aria-label="Schnell erfassen – tippen zum Schreiben, gedrückt halten zum Sprechen"
+      >
+        {isRecording ? (
+          <Mic className="w-7 h-7 stroke-[2.5]" />
+        ) : (
+          <Plus className="w-7 h-7 stroke-[2.5]" />
+        )}
+      </button>
+    </>
+  );
+};

frontend/src/features/capture/__tests__/HoldToRecordFab.test.tsx

@@ -0,0 +1,257 @@
+import React from 'react';
+import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
+import { render, screen, fireEvent, act } from '@testing-library/react';
+import { HoldToRecordFab } from '../HoldToRecordFab';
+
+class FakeSpeechRecognition {
+  static instances: FakeSpeechRecognition[] = [];
+
+  lang = '';
+  interimResults = false;
+  continuous = false;
+  started = false;
+  stopped = false;
+  aborted = false;
+
+  onresult: ((event: any) => void) | null = null;
+  onerror: ((event: any) => void) | null = null;
+  onend: (() => void) | null = null;
+
+  constructor() {
+    FakeSpeechRecognition.instances.push(this);
+  }
+
+  start() {
+    this.started = true;
+  }
+
+  stop() {
+    this.stopped = true;
+    this.onend?.();
+  }
+
+  abort() {
+    this.aborted = true;
+    this.onend?.();
+  }
+
+  emitResult(transcript: string) {
+    this.onresult?.({
+      results: [[{ transcript }]],
+    });
+  }
+
+  emitError(error: string) {
+    this.onerror?.({ error });
+  }
+}
+
+describe('HoldToRecordFab', () => {
+  beforeEach(() => {
+    vi.useFakeTimers();
+    FakeSpeechRecognition.instances = [];
+    (window as any).SpeechRecognition = FakeSpeechRecognition;
+    delete (window as any).webkitSpeechRecognition;
+  });
+
+  afterEach(() => {
+    vi.useRealTimers();
+    delete (window as any).SpeechRecognition;
+    delete (window as any).webkitSpeechRecognition;
+  });
+
+  it('calls onTap on a short tap (< 350 ms)', () => {
+    const onTap = vi.fn();
+    const onTranscript = vi.fn();
+
+    render(<HoldToRecordFab onTap={onTap} onTranscript={onTranscript} />);
+    const btn = screen.getByRole('button', {
+      name: /Schnell erfassen – tippen zum Schreiben, gedrückt halten zum Sprechen/i,
+    });
+
+    fireEvent.pointerDown(btn, { clientY: 600, button: 0 });
+    act(() => {
+      vi.advanceTimersByTime(150);
+    });
+    fireEvent.pointerUp(btn, { clientY: 600 });
+
+    expect(onTap).toHaveBeenCalledTimes(1);
+    expect(onTranscript).not.toHaveBeenCalled();
+    expect(FakeSpeechRecognition.instances).toHaveLength(0);
+  });
+
+  it('starts recording on hold (>= 350 ms) and calls onTranscript with recognized text on release', async () => {
+    const onTap = vi.fn();
+    const onTranscript = vi.fn().mockResolvedValue(undefined);
+
+    render(<HoldToRecordFab onTap={onTap} onTranscript={onTranscript} />);
+    const btn = screen.getByRole('button', {
+      name: /Schnell erfassen – tippen zum Schreiben, gedrückt halten zum Sprechen/i,
+    });
+
+    fireEvent.pointerDown(btn, { clientY: 600, button: 0 });
+    act(() => {
+      vi.advanceTimersByTime(350);
+    });
+
+    expect(FakeSpeechRecognition.instances).toHaveLength(1);
+    const rec = FakeSpeechRecognition.instances[0];
+    expect(rec.started).toBe(true);
+    expect(rec.lang).toBe('de-DE');
+    expect(rec.interimResults).toBe(true);
+    expect(rec.continuous).toBe(true);
+    expect(screen.getByText(/Ich höre zu… \(loslassen speichert\)/i)).toBeTruthy();
+
+    act(() => {
+      rec.emitResult('Milch und Kaffee einkaufen');
+    });
+    expect(screen.getByText('Milch und Kaffee einkaufen')).toBeTruthy();
+
+    await act(async () => {
+      fireEvent.pointerUp(btn, { clientY: 600 });
+    });
+
+    expect(rec.stopped).toBe(true);
+    expect(onTap).not.toHaveBeenCalled();
+    expect(onTranscript).toHaveBeenCalledTimes(1);
+    expect(onTranscript).toHaveBeenCalledWith('Milch und Kaffee einkaufen');
+    expect(screen.getByRole('status').textContent).toContain('Im Quick Dump gespeichert');
+  });
+
+  it('cancels recording when dragged up by more than 60 px and released', () => {
+    const onTap = vi.fn();
+    const onTranscript = vi.fn();
+
+    render(<HoldToRecordFab onTap={onTap} onTranscript={onTranscript} />);
+    const btn = screen.getByRole('button', {
+      name: /Schnell erfassen – tippen zum Schreiben, gedrückt halten zum Sprechen/i,
+    });
+
+    fireEvent.pointerDown(btn, { clientY: 600, button: 0 });
+    act(() => {
+      vi.advanceTimersByTime(350);
+    });
+
+    const rec = FakeSpeechRecognition.instances[0];
+    act(() => {
+      rec.emitResult('Verwerfen Test');
+    });
+
+    // Drag up by 80px (> 60px threshold)
+    fireEvent.pointerMove(btn, { clientY: 520 });
+    expect(screen.getByText(/🗑️ Zum Abbrechen loslassen/i)).toBeTruthy();
+
+    fireEvent.pointerUp(btn, { clientY: 520 });
+
+    expect(rec.aborted).toBe(true);
+    expect(onTap).not.toHaveBeenCalled();
+    expect(onTranscript).not.toHaveBeenCalled();
+  });
+
+  it('resumes normal recording if dragged above 60 px and then pulled back down before release', async () => {
+    const onTap = vi.fn();
+    const onTranscript = vi.fn().mockResolvedValue(undefined);
+
+    render(<HoldToRecordFab onTap={onTap} onTranscript={onTranscript} />);
+    const btn = screen.getByRole('button', {
+      name: /Schnell erfassen – tippen zum Schreiben, gedrückt halten zum Sprechen/i,
+    });
+
+    fireEvent.pointerDown(btn, { clientY: 600, button: 0 });
+    act(() => {
+      vi.advanceTimersByTime(350);
+    });
+
+    const rec = FakeSpeechRecognition.instances[0];
+    act(() => {
+      rec.emitResult('Doch behalten');
+    });
+
+    // Drag up into cancel zone
+    fireEvent.pointerMove(btn, { clientY: 510 });
+    expect(screen.getByText(/🗑️ Zum Abbrechen loslassen/i)).toBeTruthy();
+
+    // Drag back down below threshold
+    fireEvent.pointerMove(btn, { clientY: 580 });
+    expect(screen.getByText(/Ich höre zu… \(loslassen speichert\)/i)).toBeTruthy();
+
+    await act(async () => {
+      fireEvent.pointerUp(btn, { clientY: 580 });
+    });
+
+    expect(rec.aborted).toBe(false);
+    expect(onTranscript).toHaveBeenCalledWith('Doch behalten');
+  });
+
+  it('shows "Nichts erkannt" and does not save when transcript is empty', async () => {
+    const onTap = vi.fn();
+    const onTranscript = vi.fn();
+
+    render(<HoldToRecordFab onTap={onTap} onTranscript={onTranscript} />);
+    const btn = screen.getByRole('button', {
+      name: /Schnell erfassen – tippen zum Schreiben, gedrückt halten zum Sprechen/i,
+    });
+
+    fireEvent.pointerDown(btn, { clientY: 600, button: 0 });
+    act(() => {
+      vi.advanceTimersByTime(350);
+    });
+    await act(async () => {
+      fireEvent.pointerUp(btn, { clientY: 600 });
+    });
+
+    expect(onTap).not.toHaveBeenCalled();
+    expect(onTranscript).not.toHaveBeenCalled();
+    expect(screen.getByRole('status').textContent).toContain('Nichts erkannt');
+  });
+
+  it('falls back to onTap and shows message when microphone permission is denied', () => {
+    const onTap = vi.fn();
+    const onTranscript = vi.fn();
+
+    render(<HoldToRecordFab onTap={onTap} onTranscript={onTranscript} />);
+    const btn = screen.getByRole('button', {
+      name: /Schnell erfassen – tippen zum Schreiben, gedrückt halten zum Sprechen/i,
+    });
+
+    fireEvent.pointerDown(btn, { clientY: 600, button: 0 });
+    act(() => {
+      vi.advanceTimersByTime(350);
+    });
+
+    const rec = FakeSpeechRecognition.instances[0];
+    act(() => {
+      rec.emitError('not-allowed');
+    });
+    fireEvent.pointerUp(btn, { clientY: 600 });
+
+    expect(onTap).toHaveBeenCalledTimes(1);
+    expect(onTranscript).not.toHaveBeenCalled();
+    expect(screen.getByRole('status').textContent).toContain('Mikrofon nicht freigegeben');
+  });
+
+  it('falls back to onTap and shows a status message when SpeechRecognition is unsupported', () => {
+    delete (window as any).SpeechRecognition;
+    delete (window as any).webkitSpeechRecognition;
+
+    const onTap = vi.fn();
+    const onTranscript = vi.fn();
+
+    render(<HoldToRecordFab onTap={onTap} onTranscript={onTranscript} />);
+    const btn = screen.getByRole('button', {
+      name: /Schnell erfassen – tippen zum Schreiben, gedrückt halten zum Sprechen/i,
+    });
+
+    fireEvent.pointerDown(btn, { clientY: 600, button: 0 });
+    act(() => {
+      vi.advanceTimersByTime(350);
+    });
+    fireEvent.pointerUp(btn, { clientY: 600 });
+
+    expect(onTap).toHaveBeenCalledTimes(1);
+    expect(onTranscript).not.toHaveBeenCalled();
+    expect(screen.getByRole('status').textContent).toContain(
+      'Spracheingabe wird von diesem Browser nicht unterstützt'
+    );
+  });
+});

frontend/src/features/help/HelpView.tsx

@@ -413,7 +413,7 @@ export const HelpView: React.FC<HelpViewProps> = ({ onBack }) => {
                 <strong className="text-main">Sofortiger Fokus:</strong> Drücke irgendwo in der App <kbd className="px-1.5 py-0.5 bg-panel border border-border rounded font-mono text-[11px] text-accent">C</kbd> oder <kbd className="px-1.5 py-0.5 bg-panel border border-border rounded font-mono text-[11px] text-accent">/</kbd> (sofern du nicht bereits in einem Textfeld tippst).
               </li>
               <li>
-                <strong className="text-main">Sprach-Diktat:</strong> Klicke auf das Mikrofon-Symbol (<Mic className="w-3.5 h-3.5 inline text-accent" />) für kontinuierliche Spracheingabe über die Web Speech API. Gesprochene Gedanken werden automatisch mit dem Tag <code className="text-accent font-mono">#diktat</code> versehen.
+                <strong className="text-main">Sprach-Diktat:</strong> Klicke auf das Mikrofon-Symbol (<Mic className="w-3.5 h-3.5 inline text-accent" />) für kontinuierliche Spracheingabe über die Web Speech API. Gesprochene Gedanken werden automatisch mit dem Tag <code className="text-accent font-mono">#diktat</code> versehen. Auf dem Smartphone kannst du außerdem den runden <strong className="text-main">+</strong>-Button in der unteren Leiste gedrückt halten, um direkt einzusprechen – beim Loslassen landet der Text sofort im Posteingang (nach oben ziehen bricht ab).
               </li>
               <li>
                 <strong className="text-main">Ziel-Auswahl:</strong> Gedanken landen standardmäßig im Posteingang. Über die Optionen kannst du sie jedoch sofort als <strong className="text-main">Aufgabe</strong> (mit Priorität und Fälligkeit), <strong className="text-main">Listeneintrag</strong> (z.B. auf die Einkaufsliste) oder <strong className="text-main">neue Notiz</strong> routen.

frontend/src/test/setup.ts

@@ -4,6 +4,19 @@ import { cleanup } from '@testing-library/react';
 // jsdom has no layout engine, so scrolling is a no-op here
 Element.prototype.scrollIntoView = () => {};
 
+if (typeof window !== 'undefined' && !window.PointerEvent) {
+  class PolyfillPointerEvent extends MouseEvent {
+    pointerId: number;
+    pointerType: string;
+    constructor(type: string, params: PointerEventInit = {}) {
+      super(type, params);
+      this.pointerId = params.pointerId ?? 1;
+      this.pointerType = params.pointerType ?? 'touch';
+    }
+  }
+  (window as any).PointerEvent = PolyfillPointerEvent;
+}
+
 afterEach(() => {
   cleanup();
 });