Valgbare jingler: seks innebygde + egne filer i chimes/

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

AGENTS.md

@@ -27,7 +27,12 @@ ingen PIN), `ANNOUNCER_SINK` (pw-play `--target`; tom = standard utgang).
 - Knappen «Send ut» er låst til nøyaktig den teksten/stemmen/hastigheten som
   sist ble forhåndslyttet. Det er med vilje: ingen skal sende noe de ikke har hørt.
 - Ferdige WAV-er bufres i `cache/` etter innholdshash. Trygt å slette.
-- Chimen («ding-dong») genereres i Python, ingen lydfil.
+- Jingler: de innebygde (`BUILTIN_CHIMES`) genereres i Python med `_tone`/
+  `_mix`/`_overlay`, ingen lydfiler. Egne filer i `chimes/` (wav/mp3/ogg/flac)
+  dukker opp i menyen som `file:<stem>`; ffmpeg kutter til 6 s, normaliserer
+  til -18 LUFS og fader ut. Mappa leses ved hver forespørsel, så nye filer
+  vises uten restart. Toppnivå på de innebygde ligger mellom -4 og -10 dBFS,
+  talen går til 0 dBFS, så jinglen skal aldri være høyere enn stemmen.
 
 ## Stemmer og motorer
 

announcer.py

@@ -98,21 +98,136 @@ def load_voice(name: str):
         return _voice_cache[name]
 
 
-def make_chime(rate: int, volume: float = 0.35) -> bytes:
+# --------------------------------------------------------------------------
-    """To-tone «ding-dong» som 16-bit mono PCM i gitt samplerate."""
+# Jingler (chimes). Alle de innebygde genereres i Python; egne lydfiler
-    out = bytearray()
+# (wav/mp3/ogg/flac) i CHIME_DIR dukker opp automatisk i menyen.
-    for freq, dur in ((784.0, 0.28), (523.25, 0.42)):
+# --------------------------------------------------------------------------
+
+CHIME_DIR = Path(os.environ.get("ANNOUNCER_CHIMES", HERE / "chimes"))
+
+
+def _tone(rate: int, freq: float, dur: float, vol: float = 0.35, wave_: str = "sine",
+          decay: float = 5.5, attack: float = 0.005, harm2: float = 0.0, slide: float = 0.0) -> list[float]:
+    """Én tone som liste av floats. slide = frekvensendring (multiplikator) over tonen."""
     n = int(rate * dur)
+    out = []
+    phase = 0.0
     for i in range(n):
         t = i / rate
-            env = math.exp(-5.5 * t)  # rask utklang
+        f = freq * (slide ** (t / dur)) if slide else freq
-            s = (
+        phase += 2 * math.pi * f / rate
-                math.sin(2 * math.pi * freq * t)
+        if wave_ == "square":
-                + 0.3 * math.sin(2 * math.pi * freq * 2 * t)
+            v = 1.0 if math.sin(phase) >= 0 else -1.0
-            ) * env * volume
+        elif wave_ == "saw":
-            out += struct.pack("<h", int(max(-1.0, min(1.0, s)) * 32767))
+            v = 2.0 * ((phase / (2 * math.pi)) % 1.0) - 1.0
-    out += b"\x00\x00" * int(rate * 0.35)  # pause før talen
+        else:
-    return bytes(out)
+            v = math.sin(phase) + harm2 * math.sin(2 * phase)
+        env = min(1.0, t / attack) * math.exp(-decay * t)
+        out.append(v * env * vol)
+    return out
+
+
+def _mix(*parts: list[float], gap: float = 0.0, rate: int = 22050) -> list[float]:
+    out: list[float] = []
+    for p in parts:
+        out += p
+        out += [0.0] * int(rate * gap)
+    return out
+
+
+def _overlay(base: list[float], layer: list[float], at: float, rate: int) -> list[float]:
+    start = int(at * rate)
+    if len(base) < start + len(layer):
+        base = base + [0.0] * (start + len(layer) - len(base))
+    for i, v in enumerate(layer):
+        base[start + i] += v
+    return base
+
+
+def _dingdong(rate: int) -> list[float]:
+    return _mix(_tone(rate, 784.0, 0.28, harm2=0.3), _tone(rate, 523.25, 0.42, harm2=0.3), rate=rate)
+
+
+def _airport(rate: int) -> list[float]:
+    """Tre stigende toner med lang klang, som på en flyplass."""
+    out: list[float] = []
+    for i, f in enumerate((523.25, 659.25, 783.99)):
+        out = _overlay(out, _tone(rate, f, 1.2, vol=0.3, decay=2.2, harm2=0.2), i * 0.38, rate)
+    return out
+
+
+def _chiptune(rate: int) -> list[float]:
+    """8-bit arpeggio, som et power-up i et gammelt spill."""
+    notes = (523.25, 659.25, 783.99, 1046.5, 1318.5, 1567.98)
+    parts = [_tone(rate, f, 0.09, vol=0.3, wave_="square", decay=3.0) for f in notes]
+    parts.append(_tone(rate, 2093.0, 0.35, vol=0.3, wave_="square", decay=6.0))
+    return _mix(*parts, rate=rate)
+
+
+def _scifi(rate: int) -> list[float]:
+    """Sveip opp + to blipp, som en kommunikator."""
+    sweep = _tone(rate, 300.0, 0.35, vol=0.3, decay=1.5, slide=5.0)
+    blips = _mix(_tone(rate, 1800.0, 0.07, vol=0.25, decay=10), _tone(rate, 2400.0, 0.12, vol=0.25, decay=10), gap=0.03, rate=rate)
+    return _mix(sweep, blips, rate=rate)
+
+
+def _horn(rate: int) -> list[float]:
+    """Air horn: to lag av sagtann med litt vibrato, derav grovheten."""
+    out: list[float] = []
+    for f, v in ((233.08, 0.22), (349.23, 0.16), (466.16, 0.1)):
+        out = _overlay(out, _tone(rate, f, 0.9, vol=v, wave_="saw", decay=1.3, attack=0.03), 0.0, rate)
+    return out
+
+
+def _gong(rate: int) -> list[float]:
+    """Dyp gong med uharmoniske overtoner."""
+    out: list[float] = []
+    for f, v, d in ((110.0, 0.3, 1.0), (164.8, 0.18, 1.4), (263.0, 0.12, 2.0), (411.0, 0.07, 3.0), (650.0, 0.04, 4.0)):
+        out = _overlay(out, _tone(rate, f, 2.5, vol=v, decay=d, attack=0.004), 0.0, rate)
+    return out
+
+
+BUILTIN_CHIMES: dict[str, tuple[str, object]] = {
+    "dingdong": ("Ding-dong", _dingdong),
+    "airport": ("Flyplass (tre toner)", _airport),
+    "chiptune": ("8-bit power-up", _chiptune),
+    "scifi": ("Sci-fi kommunikator", _scifi),
+    "horn": ("Air horn", _horn),
+    "gong": ("Gong", _gong),
+}
+CHIME_EXT = {".wav", ".mp3", ".ogg", ".flac", ".m4a"}
+
+
+def file_chimes() -> dict[str, Path]:
+    if not CHIME_DIR.is_dir():
+        return {}
+    return {f"file:{p.stem}": p for p in sorted(CHIME_DIR.iterdir()) if p.suffix.lower() in CHIME_EXT}
+
+
+def list_chimes() -> list[dict]:
+    out = [{"id": "none", "name": "Ingen"}]
+    out += [{"id": k, "name": v[0]} for k, v in BUILTIN_CHIMES.items()]
+    out += [{"id": k, "name": f"{p.stem} (egen fil)"} for k, p in file_chimes().items()]
+    return out
+
+
+def chime_pcm(chime: str, rate: int) -> bytes:
+    """Jingle som 16-bit mono PCM i gitt samplerate, etterfulgt av en kort pause."""
+    if chime == "none":
+        return b""
+    if chime in BUILTIN_CHIMES:
+        samples = BUILTIN_CHIMES[chime][1](rate)
+        pcm = b"".join(struct.pack("<h", int(max(-1.0, min(1.0, v)) * 32767)) for v in samples)
+    else:
+        path = file_chimes().get(chime)
+        if path is None:
+            raise ValueError("Ukjent jingle")
+        pcm = subprocess.run(
+            ["ffmpeg", "-v", "error", "-i", str(path), "-t", "6", "-f", "s16le", "-ac", "1", "-ar", str(rate),
+             "-af", "loudnorm=I=-18,afade=t=out:st=5:d=1", "-"],
+            capture_output=True, check=True,
+        ).stdout
+    return pcm + b"\x00\x00" * int(rate * 0.35)
 
 
 def edge_pcm(text: str, voice: str, speed: float) -> bytes:
@@ -138,12 +253,12 @@ def edge_pcm(text: str, voice: str, speed: float) -> bytes:
         mp3.unlink(missing_ok=True)
 
 
-def synth_key(text: str, voice: str, speed: float, chime: bool) -> str:
+def synth_key(text: str, voice: str, speed: float, chime: str) -> str:
-    h = hashlib.sha1(f"{voice}|{speed:.2f}|{int(chime)}|{text}".encode()).hexdigest()
+    h = hashlib.sha1(f"{voice}|{speed:.2f}|{chime}|{text}".encode()).hexdigest()
     return h[:20]
 
 
-def synthesize(text: str, voice_name: str, speed: float, chime: bool) -> Path:
+def synthesize(text: str, voice_name: str, speed: float, chime: str) -> Path:
     """Returnerer sti til ferdig WAV (bufres på disk etter innhold)."""
     from piper.config import SynthesisConfig
 
@@ -156,15 +271,13 @@ def synthesize(text: str, voice_name: str, speed: float, chime: bool) -> Path:
     pcm = bytearray()
     if spec["engine"] == "edge":
         rate = EDGE_RATE
-        if chime:
+        pcm += chime_pcm(chime, rate)
-            pcm += make_chime(rate)
         pcm += edge_pcm(text, spec["model"], speed)
     else:
         voice = load_voice(spec["model"])
         rate = voice.config.sample_rate
         cfg = SynthesisConfig(length_scale=1.0 / speed, speaker_id=spec["speaker"])
-        if chime:
+        pcm += chime_pcm(chime, rate)
-            pcm += make_chime(rate)
         for chunk in voice.synthesize(text, cfg):
             pcm += chunk.audio_int16_bytes
     pcm += b"\x00\x00" * int(rate * 0.3)
@@ -190,7 +303,7 @@ class Item:
     text: str
     voice: str
     speed: float
-    chime: bool
+    chime: str
     sender: str
     created: float = field(default_factory=time.time)
     status: str = "queued"  # queued | playing | done | stopped | failed
@@ -218,7 +331,7 @@ class Player:
         self.seq = 0
         threading.Thread(target=self._loop, daemon=True, name="player").start()
 
-    def enqueue(self, text: str, voice: str, speed: float, chime: bool, sender: str) -> Item:
+    def enqueue(self, text: str, voice: str, speed: float, chime: str, sender: str) -> Item:
         with self.lock:
             self.seq += 1
             item = Item(self.seq, text, voice, speed, chime, sender)
@@ -297,7 +410,7 @@ def check_pin() -> None:
         abort(401)
 
 
-def parse_request() -> tuple[str, str, float, bool]:
+def parse_request() -> tuple[str, str, float, str]:
     data = request.get_json(silent=True) or {}
     text = " ".join(str(data.get("text", "")).split())
     if not text:
@@ -312,7 +425,14 @@ def parse_request() -> tuple[str, str, float, bool]:
     except (TypeError, ValueError):
         abort(400, "Ugyldig hastighet")
     speed = max(0.6, min(1.6, speed))
-    chime = bool(data.get("chime", True))
+    chime = data.get("chime", "dingdong")
+    if chime is True:
+        chime = "dingdong"
+    elif chime is False or chime is None:
+        chime = "none"
+    chime = str(chime)
+    if chime != "none" and chime not in BUILTIN_CHIMES and chime not in file_chimes():
+        abort(400, "Ukjent jingle")
     return text, voice, speed, chime
 
 
@@ -327,6 +447,8 @@ def config() -> Response:
         {
             "voices": [{"id": v["id"], "name": v["name"], "group": v["group"], "engine": v["engine"]} for v in VOICES_LIST],
             "default_voice": DEFAULT_VOICE,
+            "chimes": list_chimes(),
+            "default_chime": "dingdong",
             "max_chars": MAX_CHARS,
             "pin_required": bool(PIN),
         }

chimes/README.txt

@@ -0,0 +1,2 @@
+Legg egne jingler her (wav/mp3/ogg/flac). Filnavnet blir navnet i menyen.
+Maks 6 sekunder brukes, volumet normaliseres.

index.html

@@ -86,8 +86,9 @@
         <label for="speed">Hastighet <span id="speedval">1.0×</span></label>
         <input id="speed" type="range" min="0.7" max="1.4" step="0.05" value="1.0">
       </div>
-      <div style="flex:0 1 auto">
+      <div>
-        <div class="check"><input id="chime" type="checkbox" checked><label for="chime" style="margin:0;text-transform:none;letter-spacing:0;color:var(--fg)">Ding-dong først</label></div>
+        <label for="chime">Jingle først</label>
+        <select id="chime"></select>
       </div>
     </div>
 
@@ -132,7 +133,7 @@ const QUICK = [
 
 function key() { return JSON.stringify(payload()); }
 function payload() {
-  return { text: text.value.trim(), voice: voice.value, speed: parseFloat(speed.value), chime: chime.checked };
+  return { text: text.value.trim(), voice: voice.value, speed: parseFloat(speed.value), chime: chime.value };
 }
 function headers() {
   const h = { 'Content-Type': 'application/json' };
@@ -163,6 +164,11 @@ async function init() {
   }
   voice.value = localStorage.getItem('voice') || cfg.default_voice;
   if (!voice.value) voice.value = cfg.default_voice;
+  for (const c of cfg.chimes) {
+    const o = document.createElement('option'); o.value = c.id; o.textContent = c.name; chime.append(o);
+  }
+  chime.value = localStorage.getItem('chime') || cfg.default_chime;
+  if (!chime.value) chime.value = cfg.default_chime;
   $('#max').textContent = cfg.max_chars; text.maxLength = cfg.max_chars;
   if (cfg.pin_required) { $('#pinbox').style.display = 'block'; $('#pin').value = localStorage.getItem('pin') || ''; }
   for (const q of QUICK) {
@@ -176,7 +182,7 @@ async function init() {
 text.addEventListener('input', () => { $('#count').textContent = text.value.length; updateAnnounce(); });
 voice.addEventListener('change', () => { localStorage.setItem('voice', voice.value); updateAnnounce(); });
 speed.addEventListener('input', () => { $('#speedval').textContent = parseFloat(speed.value).toFixed(2).replace(/0$/, '') + '×'; updateAnnounce(); });
-chime.addEventListener('change', updateAnnounce);
+chime.addEventListener('change', () => { localStorage.setItem('chime', chime.value); updateAnnounce(); });
 $('#pin').addEventListener('input', e => localStorage.setItem('pin', e.target.value));
 
 btnPreview.addEventListener('click', async () => {