Hitonabi/mission-control-v2
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Feat: Edge-TTS-Engine (native dt. Azure-Stimmen, kein Akzent) + Probe-hören-Knopf

Browse Source 
Edge-TTS als 4. Engine (gratis, kein Key, KEIN Cloning → natives Deutsch ohne Akzent — die einzige
Lösung gegen das Akzent-Problem aller Cloning-Engines). /voices listet dt. Edge-Stimmen (weiblich
zuerst, inkl. Gisela). Picker: Engine 'Edge (natürlich · gratis)' + Probe-hören-Knopf (festen Satz je
Engine/Stimme abspielen, ohne reinsprechen). Default bleibt Piper.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Hitonabi
2026-06-28 00:44:01 +02:00
parent aa62c98247
commit 11f0066b47
7 changed files with 257 additions and 154 deletions
Download Patch File Download Diff File Expand all files Collapse all files
frontend/dist/assets/GraphView-DFF0zsBc.js → frontend/dist/assets/GraphView-BlR6l9-5.js
Vendored
+1 -1
View File
File diff suppressed because one or more lines are too long
frontend/dist/assets/index-DzmNpOtE.js → frontend/dist/assets/index-DesgNIBq.js
Vendored
+141 -141
View File
File diff suppressed because one or more lines are too long
frontend/dist/assets/index-DueAT_Qc.css → frontend/dist/assets/index-jNLzkOWh.css
Vendored
+1 -1
View File
File diff suppressed because one or more lines are too long
frontend/dist/index.html
Vendored
+2 -2
View File
@@ -7,8 +7,8 @@
<link rel="manifest" href="/manifest.webmanifest" /> <link rel="manifest" href="/manifest.webmanifest" />
<link rel="icon" type="image/svg+xml" href="/favicon.svg" /> <link rel="icon" type="image/svg+xml" href="/favicon.svg" />
<title>Mission Control 2.0</title> <title>Mission Control 2.0</title>
<script type="module" crossorigin src="/assets/index-DzmNpOtE.js"></script> <script type="module" crossorigin src="/assets/index-DesgNIBq.js"></script>
<link rel="stylesheet" crossorigin href="/assets/index-DueAT_Qc.css"> <link rel="stylesheet" crossorigin href="/assets/index-jNLzkOWh.css">
</head> </head>
<body> <body>
<div id="root"></div> <div id="root"></div>
frontend/src/components/voice/AvatarPicker.tsx
+48 -4
View File
@@ -1,5 +1,5 @@
import { useEffect, useRef, useState } from "react" import { useEffect, useRef, useState } from "react"
import { Upload, Link2, ExternalLink, Check, Sparkles } from "lucide-react" import { Upload, Link2, ExternalLink, Check, Sparkles, Volume2, Loader2 } from "lucide-react"
import { saveUploadedVrm, loadUploadedVrm } from "@/lib/voice/vrmStore" import { saveUploadedVrm, loadUploadedVrm } from "@/lib/voice/vrmStore"
// Avatar selbst aussuchen: kuratierte Galerie (öffentliche, CORS-freie VRMs) + eigenes .vrm // Avatar selbst aussuchen: kuratierte Galerie (öffentliche, CORS-freie VRMs) + eigenes .vrm
@@ -26,7 +26,36 @@ export function AvatarPicker({ avatarUrl, onAvatarChange }: {
const [voices, setVoices] = useState<Voice[]>([]) const [voices, setVoices] = useState<Voice[]>([])
const [engine, setEngine] = useState(localStorage.getItem("mc_voice_engine") || "piper") const [engine, setEngine] = useState(localStorage.getItem("mc_voice_engine") || "piper")
const [voice, setVoice] = useState(localStorage.getItem("mc_voice_voice") || "") const [voice, setVoice] = useState(localStorage.getItem("mc_voice_voice") || "")
const [previewing, setPreviewing] = useState(false)
const fileRef = useRef<HTMLInputElement>(null) const fileRef = useRef<HTMLInputElement>(null)
const previewAudio = useRef<HTMLAudioElement | null>(null)
// „Probe hören": synthetisiert einen festen Satz mit der aktuellen Engine+Stimme und spielt ihn ab.
const playPreview = async () => {
if (previewing) return
setPreviewing(true)
try {
const r = await fetch("/api/voice/tts", {
method: "POST",
headers: { "Content-Type": "application/json" },
body: JSON.stringify({
text: "Hallo! Ich bin deine Assistentin. So klingt diese Stimme auf Deutsch.",
engine, voice,
}),
})
if (!r.ok) throw new Error(`TTS ${r.status}`)
const url = URL.createObjectURL(await r.blob())
previewAudio.current?.pause()
const a = new Audio(url)
previewAudio.current = a
a.onended = () => URL.revokeObjectURL(url)
await a.play()
} catch (e) {
console.error("Probe fehlgeschlagen:", e)
} finally {
setPreviewing(false)
}
}
// Hochgeladenes VRM nach Reload wiederherstellen. // Hochgeladenes VRM nach Reload wiederherstellen.
useEffect(() => { useEffect(() => {
@@ -70,8 +99,9 @@ export function AvatarPicker({ avatarUrl, onAvatarChange }: {
localStorage.setItem("mc_voice_voice", v) localStorage.setItem("mc_voice_voice", v)
} }
// Feste Reihenfolge der bekannten Engines (ElevenLabs auch ohne Key sichtbar → „Key fehlt"-Hinweis). // Feste Reihenfolge der bekannten Engines (Edge zuerst = natürlich+gratis; ElevenLabs auch ohne Key
const enginesAvail = ["piper", "chatterbox", "elevenlabs"] // sichtbar → „Key fehlt"-Hinweis).
const enginesAvail = ["edge", "piper", "chatterbox", "elevenlabs"]
const voicesForEngine = voices.filter((v) => v.engine === engine) const voicesForEngine = voices.filter((v) => v.engine === engine)
return ( return (
@@ -158,7 +188,7 @@ export function AvatarPicker({ avatarUrl, onAvatarChange }: {
engine === eng ? "border-primary/50 bg-primary/10 text-primary" : "border-border/40 hover:bg-accent" engine === eng ? "border-primary/50 bg-primary/10 text-primary" : "border-border/40 hover:bg-accent"
}`} }`}
> >
{eng === "piper" ? "Piper (schnell)" : eng === "chatterbox" ? "Chatterbox (lokal)" : eng === "elevenlabs" ? "ElevenLabs (premium)" : eng} {eng === "edge" ? "Edge (natürlich · gratis)" : eng === "piper" ? "Piper (lokal)" : eng === "chatterbox" ? "Chatterbox (lokal)" : eng === "elevenlabs" ? "ElevenLabs (premium)" : eng}
</button> </button>
))} ))}
</div> </div>
@@ -172,6 +202,20 @@ export function AvatarPicker({ avatarUrl, onAvatarChange }: {
<option key={v.id} value={v.id}>{v.label}{v.clonable ? " · klonbar" : ""}</option> <option key={v.id} value={v.id}>{v.label}{v.clonable ? " · klonbar" : ""}</option>
))} ))}
</select> </select>
<button
onClick={playPreview}
disabled={previewing}
className="flex w-full items-center justify-center gap-2 rounded-md border border-primary/40 bg-primary/10 px-2.5 py-2 text-xs text-primary hover:bg-primary/20 transition-colors disabled:opacity-60"
>
{previewing ? <Loader2 className="h-3.5 w-3.5 animate-spin" /> : <Volume2 className="h-3.5 w-3.5" />}
{previewing ? "Spielt …" : "Probe hören"}
</button>
{engine === "edge" && (
<p className="text-[11px] text-muted-foreground">
Edge-TTS: native deutsche Azure-Stimmen — sehr natürlich, kein Akzent, gratis & ohne Key.
Jüngste Stimme: <code>Gisela</code>. Cloud (Text → Microsoft).
</p>
)}
{engine === "chatterbox" && ( {engine === "chatterbox" && (
<p className="text-[11px] text-muted-foreground"> <p className="text-[11px] text-muted-foreground">
Chatterbox läuft auf CPU → erste Antwort kann ein paar Sekunden dauern. Lokal + Voice-Cloning (dt. mit Akzent). Chatterbox läuft auf CPU → erste Antwort kann ein paar Sekunden dauern. Lokal + Voice-Cloning (dt. mit Akzent).
voice_service/app.py
+60 -2
View File
@@ -60,6 +60,8 @@ CHATTERBOX_REF = os.environ.get("VOICE_CHATTERBOX_REF", "")
ELEVEN_MODEL = os.environ.get("VOICE_ELEVENLABS_MODEL", "eleven_flash_v2_5") ELEVEN_MODEL = os.environ.get("VOICE_ELEVENLABS_MODEL", "eleven_flash_v2_5")
ELEVEN_DEFAULT_VOICE = os.environ.get("VOICE_ELEVENLABS_VOICE", "21m00Tcm4TlvDq8ikWAM") ELEVEN_DEFAULT_VOICE = os.environ.get("VOICE_ELEVENLABS_VOICE", "21m00Tcm4TlvDq8ikWAM")
HERMES_ENV = Path.home() / ".hermes" / ".env" HERMES_ENV = Path.home() / ".hermes" / ".env"
# Edge-TTS: native deutsche Azure-Neural-Stimmen, gratis & ohne Key (kein Cloning → kein Akzent).
EDGE_DEFAULT = os.environ.get("VOICE_EDGE_VOICE", "de-DE-KatjaNeural")
def _eleven_key() -> str: def _eleven_key() -> str:
@@ -230,6 +232,60 @@ def elevenlabs_list() -> list[dict]:
return [] return []
# =================================================================================
# TTS — Edge-TTS (native dt. Azure-Neural-Stimmen, gratis, kein Key, KEIN Cloning → kein Akzent)
# =================================================================================
def edge_tts_synth(text: str, voice: str) -> bytes:
import asyncio
import edge_tts
v = voice or EDGE_DEFAULT
async def _run() -> bytes:
buf = bytearray()
communicate = edge_tts.Communicate(text, v)
async for chunk in communicate.stream():
if chunk["type"] == "audio":
buf.extend(chunk["data"])
return bytes(buf)
try:
return asyncio.run(_run())
except Exception as exc: # noqa: BLE001
raise HTTPException(502, f"Edge-TTS-Fehler: {exc}")
_edge_cache: list | None = None
def edge_list() -> list[dict]:
"""Deutsche Edge/Azure-Stimmen (einmal vom Dienst geholt + gecacht). Weibliche zuerst."""
global _edge_cache
if _edge_cache is not None:
return _edge_cache
try:
import asyncio
import edge_tts
voices = asyncio.run(edge_tts.list_voices())
out = []
for v in voices:
if not v.get("Locale", "").startswith("de-"):
continue
short = v["ShortName"] # z.B. de-DE-KatjaNeural
gender = "weiblich" if v.get("Gender") == "Female" else "männlich"
name = short.split("-")[-1].replace("Neural", "").replace("Multilingual", " (multiling.)")
out.append({"engine": "edge", "id": short,
"label": f"{name} · {v['Locale']} · {gender}",
"clonable": False, "_female": v.get("Gender") == "Female"})
out.sort(key=lambda i: (not i.pop("_female"), i["id"]))
_edge_cache = out
except Exception: # noqa: BLE001
log.warning("Edge-Stimmenliste fehlgeschlagen", exc_info=True)
_edge_cache = []
return _edge_cache
# ================================================================================= # =================================================================================
# FastAPI # FastAPI
# ================================================================================= # =================================================================================
@@ -258,14 +314,14 @@ class TTSIn(BaseModel):
@app.get("/health") @app.get("/health")
def health() -> dict: def health() -> dict:
return {"ok": True, "stt_model": STT_MODEL, return {"ok": True, "stt_model": STT_MODEL,
"engines": ["piper", "chatterbox", "elevenlabs"], "engines": ["edge", "piper", "chatterbox", "elevenlabs"],
"elevenlabs_key": bool(_eleven_key()), "elevenlabs_key": bool(_eleven_key()),
"piper_voices": [v["id"] for v in piper_list()]} "piper_voices": [v["id"] for v in piper_list()]}
@app.get("/voices") @app.get("/voices")
def voices() -> dict: def voices() -> dict:
return {"voices": piper_list() + chatterbox_list() + elevenlabs_list(), return {"voices": edge_list() + piper_list() + chatterbox_list() + elevenlabs_list(),
"default": {"engine": "piper", "voice": PIPER_DEFAULT}} "default": {"engine": "piper", "voice": PIPER_DEFAULT}}
@@ -284,6 +340,8 @@ def tts(body: TTSIn) -> Response:
text = (body.text or "").strip() text = (body.text or "").strip()
if not text: if not text:
raise HTTPException(400, "Leerer Text.") raise HTTPException(400, "Leerer Text.")
if body.engine == "edge":
return Response(content=edge_tts_synth(text, body.voice), media_type="audio/mpeg")
if body.engine == "elevenlabs": if body.engine == "elevenlabs":
return Response(content=elevenlabs_tts(text, body.voice), media_type="audio/mpeg") return Response(content=elevenlabs_tts(text, body.voice), media_type="audio/mpeg")
if body.engine == "chatterbox": if body.engine == "chatterbox":
voice_service/requirements.txt
+4 -3
View File
@@ -1,9 +1,10 @@
# Kern (immer nötig — STT + Piper-TTS + Server). Chatterbox + torch zieht install.sh # Kern (immer nötig — STT + Server). Piper-TTS läuft über das offizielle Binary (install.sh lädt
# separat best-effort nach (CPU-Wheel), damit der Voice-Loop auch ohne Premium-Engine läuft. # es), nicht über das PyPI-Paket (piper-phonemize hat auf neueren Linux keine Wheels). Chatterbox
# + torch zieht install.sh best-effort nach (CPU-Wheel), damit der Loop auch ohne Premium läuft.
fastapi fastapi
uvicorn uvicorn
python-multipart python-multipart
soundfile soundfile
numpy numpy
faster-whisper faster-whisper
piper-tts==1.2.0 edge-tts