mirror of
https://github.com/OpenSquawk/OpenSquawk
synced 2026-08-05 00:46:00 +08:00
iwas was nich tut dazutun
This commit is contained in:
File diff suppressed because it is too large
Load Diff
287
app/pages/say.vue
Normal file
287
app/pages/say.vue
Normal file
@@ -0,0 +1,287 @@
|
||||
<template>
|
||||
<v-container class="py-8" max-width="900">
|
||||
<v-row class="mb-4" align="center" justify="space-between">
|
||||
<h1 class="text-h5">ATC Generator</h1>
|
||||
<v-btn variant="outlined" prepend-icon="mdi-broom" @click="resetForm">Zurücksetzen</v-btn>
|
||||
</v-row>
|
||||
|
||||
<v-card class="pa-4 mb-6">
|
||||
<v-row dense>
|
||||
<v-col cols="12">
|
||||
<v-textarea
|
||||
v-model="form.text"
|
||||
label="ATC-Text (z. B. „D-ABCD ready for taxi…“)"
|
||||
rows="4"
|
||||
auto-grow
|
||||
:counter="500"
|
||||
required
|
||||
/>
|
||||
</v-col>
|
||||
|
||||
<v-col cols="12" md="6">
|
||||
<v-slider
|
||||
v-model="form.level"
|
||||
:min="1" :max="5" :step="1"
|
||||
label="Radio-Level (1–5)"
|
||||
show-ticks tick-size="3" thumb-label
|
||||
/>
|
||||
<div class="text-caption mt-1">
|
||||
1 = stark verrauscht · 4 = gut (Default) · 5 = klar
|
||||
</div>
|
||||
</v-col>
|
||||
|
||||
<v-col cols="12" md="3">
|
||||
<v-select
|
||||
v-model="form.voice"
|
||||
:items="voices"
|
||||
label="Stimme"
|
||||
item-title="label" item-value="value"
|
||||
/>
|
||||
</v-col>
|
||||
|
||||
<v-col cols="12" md="3">
|
||||
<v-text-field v-model="form.tag" label="Tag (optional)" />
|
||||
</v-col>
|
||||
|
||||
<v-col cols="12" class="d-flex gap-3">
|
||||
<v-btn color="primary" :loading="loading" @click="generate">
|
||||
<v-icon start>mdi-play</v-icon> Erzeugen & Abspielen
|
||||
</v-btn>
|
||||
<v-btn variant="tonal" :disabled="!audioUrl" @click="download">
|
||||
<v-icon start>mdi-download</v-icon> OGG herunterladen
|
||||
</v-btn>
|
||||
<v-btn variant="text" :disabled="!result" @click="copyMeta">
|
||||
<v-icon start>mdi-content-copy</v-icon> Meta kopieren
|
||||
</v-btn>
|
||||
<v-spacer/>
|
||||
<v-switch v-model="autoPlay" inset label="Auto-Play" hide-details/>
|
||||
<v-switch v-model="persistHistory" inset label="History speichern" hide-details/>
|
||||
</v-col>
|
||||
</v-row>
|
||||
</v-card>
|
||||
|
||||
<v-card class="pa-4 mb-6" v-if="result">
|
||||
<v-row>
|
||||
<v-col cols="12" md="6">
|
||||
<audio ref="player" :src="audioUrl || undefined" controls style="width:100%"></audio>
|
||||
<div class="text-caption mt-2">MIME: audio/ogg; codecs=opus · 16 kHz mono</div>
|
||||
</v-col>
|
||||
<v-col cols="12" md="6">
|
||||
<v-list density="compact">
|
||||
<v-list-item title="ID" :subtitle="result.id"/>
|
||||
<v-list-item title="Erstellt" :subtitle="new Date(result.createdAt || Date.now()).toLocaleString()"/>
|
||||
<v-list-item title="Level" :subtitle="String(result.level)"/>
|
||||
<v-list-item title="Voice" :subtitle="result.voice"/>
|
||||
<v-list-item title="Eingabe" :subtitle="result.text"/>
|
||||
<v-list-item title="Normalisiert" :subtitle="result.normalized"/>
|
||||
<v-list-item v-if="result.tag" title="Tag" :subtitle="result.tag"/>
|
||||
</v-list>
|
||||
</v-col>
|
||||
</v-row>
|
||||
</v-card>
|
||||
|
||||
<v-expand-transition>
|
||||
<v-card v-if="history.length" class="pa-4">
|
||||
<div class="d-flex align-center justify-space-between mb-2">
|
||||
<h2 class="text-subtitle-1">History</h2>
|
||||
<v-btn size="small" variant="text" @click="clearHistory"><v-icon>mdi-delete</v-icon> leeren</v-btn>
|
||||
</div>
|
||||
<v-table density="compact">
|
||||
<thead>
|
||||
<tr>
|
||||
<th>Datum</th><th>Level</th><th>Voice</th><th>Text</th><th>Play</th>
|
||||
</tr>
|
||||
</thead>
|
||||
<tbody>
|
||||
<tr v-for="h in history" :key="h.id">
|
||||
<td>{{ new Date(h.createdAt).toLocaleString() }}</td>
|
||||
<td>{{ h.level }}</td>
|
||||
<td>{{ h.voice }}</td>
|
||||
<td class="truncate">{{ h.text }}</td>
|
||||
<td>
|
||||
<v-btn size="small" @click="playFromHistory(h)"><v-icon>mdi-play</v-icon></v-btn>
|
||||
</td>
|
||||
</tr>
|
||||
</tbody>
|
||||
</v-table>
|
||||
</v-card>
|
||||
</v-expand-transition>
|
||||
|
||||
<v-snackbar v-model="snack.show" :timeout="3000" color="error">{{ snack.msg }}</v-snackbar>
|
||||
</v-container>
|
||||
</template>
|
||||
|
||||
<script setup lang="ts">
|
||||
import { ref, onMounted } from 'vue'
|
||||
|
||||
type Json = any
|
||||
|
||||
const form = ref({ text: '', level: 4, voice: 'alloy', tag: '' })
|
||||
const loading = ref(false)
|
||||
const result = ref<Json | null>(null)
|
||||
const rawJson = ref<Json | null>(null)
|
||||
const audioUrl = ref<string | null>(null)
|
||||
const player = ref<HTMLAudioElement | null>(null)
|
||||
const autoPlay = ref(true)
|
||||
const persistHistory = ref(true)
|
||||
const snack = ref({ show: false, msg: '' })
|
||||
const debugMsg = ref<string>('')
|
||||
|
||||
function err(msg: string) { snack.value = { show: true, msg }; debugMsg.value = msg }
|
||||
const nextTick = (fn: () => void) => Promise.resolve().then(fn)
|
||||
|
||||
// --- Robust: erste Base64 + zugehöriges MIME irgendwo im Objekt finden ---
|
||||
function findFirstAudio(o: Json): { base64: string; mime: string } | null {
|
||||
let found: { base64: string; mime?: string } | null = null
|
||||
function walk(x: Json, path: string[] = []) {
|
||||
if (!x || found) return
|
||||
if (typeof x === 'object') {
|
||||
// Kandidat: {base64, mime}
|
||||
if (typeof x.base64 === 'string') {
|
||||
found = { base64: x.base64, mime: typeof x.mime === 'string' ? x.mime : undefined }
|
||||
}
|
||||
// tiefer gehen
|
||||
for (const k of Object.keys(x)) walk(x[k], path.concat(k))
|
||||
}
|
||||
}
|
||||
walk(o)
|
||||
if (!found) return null
|
||||
const mime = found.mime || guessMimeFromObject(o) || 'audio/ogg'
|
||||
return { base64: found.base64, mime }
|
||||
}
|
||||
|
||||
function guessMimeFromObject(o: Json): string | null {
|
||||
// Häufige Felder prüfen
|
||||
if (o?.audio?.radio?.mime) return o.audio.radio.mime
|
||||
if (o?.audio?.mime) return o.audio.mime
|
||||
if (o?.audio?.clean?.mime) return o.audio.clean.mime
|
||||
return null
|
||||
}
|
||||
|
||||
// --- Base64 normalisieren ---
|
||||
function normalizeBase64(b64: string) {
|
||||
if (!b64) throw new Error('leer')
|
||||
// dataURL → payload
|
||||
const i = b64.indexOf(',')
|
||||
if (i !== -1 && b64.slice(0, 40).includes('base64')) b64 = b64.slice(i + 1)
|
||||
b64 = b64.replace(/\s+/g, '').replace(/-/g, '+').replace(/_/g, '/')
|
||||
const mod = b64.length % 4
|
||||
if (mod === 2) b64 += '=='
|
||||
else if (mod === 3) b64 += '='
|
||||
else if (mod === 1) throw new Error('Base64-Länge ungültig')
|
||||
return b64
|
||||
}
|
||||
|
||||
function base64ToBlobUrl(b64: string, mime = 'application/octet-stream') {
|
||||
const clean = normalizeBase64(b64)
|
||||
const bin = atob(clean)
|
||||
const bytes = new Uint8Array(bin.length)
|
||||
for (let i = 0; i < bin.length; i++) bytes[i] = bin.charCodeAt(i)
|
||||
const blob = new Blob([bytes], { type: mime })
|
||||
return URL.createObjectURL(blob)
|
||||
}
|
||||
|
||||
async function generate() {
|
||||
const text = form.value.text.trim()
|
||||
if (!text) return err('Text fehlt.')
|
||||
|
||||
loading.value = true
|
||||
debugMsg.value = ''
|
||||
try {
|
||||
const res = await $fetch<Json>('/api/atc/generate', {
|
||||
method: 'POST',
|
||||
body: { text, level: form.value.level, voice: form.value.voice, tag: form.value.tag || undefined }
|
||||
})
|
||||
rawJson.value = res
|
||||
console.log('ATC response:', res)
|
||||
|
||||
// kompatible Anzeigenamen
|
||||
const displayText = res.text ?? res.atcText ?? text
|
||||
result.value = { ...res, text: displayText, createdAt: res.createdAt ?? new Date().toISOString() }
|
||||
|
||||
// Audio extrahieren (egal wo)
|
||||
const audio = findFirstAudio(res)
|
||||
if (!audio) {
|
||||
err('Kein Audio in Response gefunden.')
|
||||
console.warn('Response ohne erkennbares Audio:', res)
|
||||
audioUrl.value = null
|
||||
return
|
||||
}
|
||||
|
||||
// Blob-URL bauen
|
||||
if (audioUrl.value) URL.revokeObjectURL(audioUrl.value)
|
||||
audioUrl.value = base64ToBlobUrl(audio.base64, audio.mime)
|
||||
|
||||
if (autoPlay.value) nextTick(() => player.value?.play().catch(() => {}))
|
||||
if (persistHistory.value) pushHistory(result.value)
|
||||
debugMsg.value = `Audio OK · mime=${audio.mime} · base64Len=${audio.base64.length}`
|
||||
} catch (e: any) {
|
||||
console.error(e)
|
||||
err(e?.data?.message || e?.message || 'Fehler beim Generieren')
|
||||
} finally {
|
||||
loading.value = false
|
||||
}
|
||||
}
|
||||
|
||||
function download() {
|
||||
if (!audioUrl.value || !result.value) return
|
||||
const mime =
|
||||
result.value?.audio?.radio?.mime ||
|
||||
result.value?.audio?.mime ||
|
||||
result.value?.audio?.clean?.mime ||
|
||||
'audio/ogg'
|
||||
const ext = mime.includes('ogg') ? 'ogg' : mime.includes('wav') ? 'wav' : 'bin'
|
||||
const a = document.createElement('a')
|
||||
a.href = audioUrl.value
|
||||
a.download = `${result.value?.id || 'atc'}.${ext}`
|
||||
a.click()
|
||||
}
|
||||
|
||||
async function copyMeta() {
|
||||
if (!result.value) return
|
||||
const meta = {
|
||||
id: result.value.id ?? null,
|
||||
createdAt: result.value.createdAt,
|
||||
level: result.value.level,
|
||||
voice: result.value.voice,
|
||||
text: result.value.text ?? '',
|
||||
normalized: result.value.normalized ?? null,
|
||||
stored: result.value.stored ?? null
|
||||
}
|
||||
await navigator.clipboard.writeText(JSON.stringify(meta, null, 2))
|
||||
}
|
||||
|
||||
const history = ref<Json[]>([])
|
||||
function pushHistory(e: Json) {
|
||||
const list = [e, ...history.value].slice(0, 50)
|
||||
history.value = list
|
||||
localStorage.setItem('atc-history', JSON.stringify(list))
|
||||
}
|
||||
function loadHistory() {
|
||||
const raw = localStorage.getItem('atc-history')
|
||||
if (raw) { try { history.value = JSON.parse(raw) } catch {} }
|
||||
}
|
||||
function clearHistory() { history.value = []; localStorage.removeItem('atc-history') }
|
||||
function playFromHistory(h: Json) {
|
||||
try {
|
||||
const audio = findFirstAudio(h)
|
||||
if (!audio) return err('History-Eintrag ohne Audio.')
|
||||
if (audioUrl.value) URL.revokeObjectURL(audioUrl.value)
|
||||
audioUrl.value = base64ToBlobUrl(audio.base64, audio.mime)
|
||||
result.value = h
|
||||
nextTick(() => player.value?.play().catch(() => {}))
|
||||
} catch (e: any) { err(e.message) }
|
||||
}
|
||||
|
||||
onMounted(loadHistory)
|
||||
</script>
|
||||
|
||||
|
||||
<style scoped>
|
||||
.truncate {
|
||||
max-width: 420px;
|
||||
white-space: nowrap;
|
||||
overflow: hidden;
|
||||
text-overflow: ellipsis;
|
||||
}
|
||||
</style>
|
||||
142
composables/radioTts.ts
Normal file
142
composables/radioTts.ts
Normal file
@@ -0,0 +1,142 @@
|
||||
// audioRadioTTS.ts
|
||||
/** AUDIO: English TTS with radio ambience (noise + bandlimit + PTT click) **/
|
||||
declare const cfg:
|
||||
| { value?: { tts?: boolean } }
|
||||
| undefined; // optional global (Nuxt/Pinia o.ä.)
|
||||
|
||||
let _ctx: AudioContext | null = null;
|
||||
let _masterGain: GainNode;
|
||||
let _noiseGain: GainNode;
|
||||
let _noiseFilterBand: BiquadFilterNode;
|
||||
let _noiseFilterHP: BiquadFilterNode;
|
||||
let _noiseFilterLP: BiquadFilterNode;
|
||||
let _compressor: DynamicsCompressorNode;
|
||||
|
||||
function ensureAudioGraph() {
|
||||
if (_ctx) return;
|
||||
_ctx = new (window.AudioContext || (window as any).webkitAudioContext)();
|
||||
|
||||
_masterGain = _ctx.createGain();
|
||||
_masterGain.gain.value = 0.9;
|
||||
|
||||
// Background "radio" noise
|
||||
const noiseBuffer = _ctx.createBuffer(1, _ctx.sampleRate * 2, _ctx.sampleRate);
|
||||
const data = noiseBuffer.getChannelData(0);
|
||||
for (let i = 0; i < data.length; i++) {
|
||||
const w = Math.random() * 2 - 1;
|
||||
data[i] = (data[i - 1] || 0) * 0.97 + w * 0.03; // cheap pink-ish
|
||||
}
|
||||
const noise = _ctx.createBufferSource();
|
||||
noise.buffer = noiseBuffer;
|
||||
noise.loop = true;
|
||||
|
||||
_noiseFilterHP = _ctx.createBiquadFilter();
|
||||
_noiseFilterHP.type = "highpass";
|
||||
_noiseFilterHP.frequency.value = 300;
|
||||
|
||||
_noiseFilterBand = _ctx.createBiquadFilter();
|
||||
_noiseFilterBand.type = "bandpass";
|
||||
_noiseFilterBand.frequency.value = 1600;
|
||||
_noiseFilterBand.Q.value = 0.6;
|
||||
|
||||
_noiseFilterLP = _ctx.createBiquadFilter();
|
||||
_noiseFilterLP.type = "lowpass";
|
||||
_noiseFilterLP.frequency.value = 3200;
|
||||
|
||||
_compressor = _ctx.createDynamicsCompressor();
|
||||
_compressor.threshold.value = -24;
|
||||
_compressor.knee.value = 20;
|
||||
_compressor.ratio.value = 6;
|
||||
_compressor.attack.value = 0.003;
|
||||
_compressor.release.value = 0.1;
|
||||
|
||||
_noiseGain = _ctx.createGain();
|
||||
_noiseGain.gain.value = 0.07;
|
||||
|
||||
noise
|
||||
.connect(_noiseFilterHP)
|
||||
.connect(_noiseFilterBand)
|
||||
.connect(_noiseFilterLP)
|
||||
.connect(_noiseGain)
|
||||
.connect(_compressor)
|
||||
.connect(_masterGain)
|
||||
.connect(_ctx.destination);
|
||||
|
||||
noise.start();
|
||||
}
|
||||
|
||||
function pttClick(durationMs = 25, gain = 0.5) {
|
||||
if (!_ctx) return;
|
||||
const osc = _ctx.createOscillator();
|
||||
const g = _ctx.createGain();
|
||||
osc.type = "square";
|
||||
osc.frequency.value = 1200 + Math.random() * 600;
|
||||
g.gain.value = gain;
|
||||
osc.connect(g).connect(_masterGain);
|
||||
osc.start();
|
||||
const now = _ctx.currentTime;
|
||||
g.gain.setValueAtTime(gain, now);
|
||||
g.gain.exponentialRampToValueAtTime(0.001, now + durationMs / 1000);
|
||||
osc.stop(now + durationMs / 1000 + 0.02);
|
||||
}
|
||||
|
||||
/** Helpers */
|
||||
function isTtsEnabled(): boolean {
|
||||
// Wenn cfg fehlt → default: true (nicht crashen, trotzdem sprechen)
|
||||
try {
|
||||
// @ts-ignore – cfg evtl. global
|
||||
const val = (cfg as any)?.value?.tts;
|
||||
return val ?? true;
|
||||
} catch {
|
||||
return true;
|
||||
}
|
||||
}
|
||||
|
||||
/** Main speak with radio ambience */
|
||||
export default function speak(text: string) {
|
||||
if (typeof window === "undefined") return;
|
||||
if (!("speechSynthesis" in window)) return;
|
||||
if (!isTtsEnabled()) return;
|
||||
|
||||
ensureAudioGraph();
|
||||
|
||||
const u = new SpeechSynthesisUtterance(text);
|
||||
u.lang = "en-US"; // or "en-GB"
|
||||
u.rate = 1.0;
|
||||
u.pitch = 1.0;
|
||||
|
||||
u.onstart = () => {
|
||||
_ctx?.resume();
|
||||
pttClick(18, 0.35);
|
||||
if (_noiseGain) {
|
||||
const now = _ctx!.currentTime;
|
||||
_noiseGain.gain.cancelScheduledValues(now);
|
||||
_noiseGain.gain.setValueAtTime(_noiseGain.gain.value, now);
|
||||
_noiseGain.gain.linearRampToValueAtTime(0.1, now + 0.05);
|
||||
}
|
||||
};
|
||||
|
||||
u.onend = () => {
|
||||
pttClick(22, 0.28);
|
||||
if (_noiseGain) {
|
||||
const now = _ctx!.currentTime;
|
||||
_noiseGain.gain.cancelScheduledValues(now);
|
||||
_noiseGain.gain.setValueAtTime(_noiseGain.gain.value, now);
|
||||
_noiseGain.gain.linearRampToValueAtTime(0.07, now + 0.2);
|
||||
}
|
||||
};
|
||||
|
||||
setTimeout(() => {
|
||||
window.speechSynthesis.cancel();
|
||||
window.speechSynthesis.speak(u);
|
||||
}, 60);
|
||||
}
|
||||
|
||||
|
||||
export function createRadioTTS(isEnabled: () => boolean = () => true) {
|
||||
return (text: string) => {
|
||||
if (!isEnabled()) return;
|
||||
speak(text);
|
||||
};
|
||||
}
|
||||
// usage: const speakRadio = createRadioTTS(() => cfg?.value?.tts ?? true);
|
||||
43
composables/radioTtsNew.ts
Normal file
43
composables/radioTtsNew.ts
Normal file
@@ -0,0 +1,43 @@
|
||||
export type RadioOpts = {
|
||||
level?: number; // 1..5
|
||||
voice?: string; // z.B. "alloy"
|
||||
};
|
||||
|
||||
export default function useRadioTTS() {
|
||||
let audio: HTMLAudioElement | null = null;
|
||||
|
||||
function stop() {
|
||||
if (audio) {
|
||||
audio.pause();
|
||||
audio.src = "";
|
||||
audio.load();
|
||||
audio = null;
|
||||
}
|
||||
}
|
||||
|
||||
function speakBrowser(text: string) {
|
||||
stop();
|
||||
const u = new SpeechSynthesisUtterance(text);
|
||||
u.lang = "en-US";
|
||||
u.rate = 1;
|
||||
u.pitch = 1;
|
||||
window.speechSynthesis.cancel();
|
||||
window.speechSynthesis.speak(u);
|
||||
return u;
|
||||
}
|
||||
|
||||
function speakServer(text: string, opts: RadioOpts = {}) {
|
||||
stop();
|
||||
const params = new URLSearchParams({
|
||||
text,
|
||||
level: String(opts.level ?? 4),
|
||||
voice: String(opts.voice ?? "alloy"),
|
||||
});
|
||||
audio = new Audio(`/api/atc/say.stream?${params.toString()}`);
|
||||
audio.autoplay = true;
|
||||
audio.play().catch(() => {/* ignore */});
|
||||
return audio;
|
||||
}
|
||||
|
||||
return { speakBrowser, speakServer, stop };
|
||||
}
|
||||
1
out.ogg
Normal file
1
out.ogg
Normal file
@@ -0,0 +1 @@
|
||||
{"error":true,"message":"[wav @ 0x14e704c40] invalid start code [255][243][228][196] in RIFF header\n[in#0 @ 0x600001370900] Error opening input: Invalid data found when processing input\nError opening input file pipe:0.\nError opening input files: Invalid data found when processing input\n"}
|
||||
199
server/api/atc/say.post.ts
Normal file
199
server/api/atc/say.post.ts
Normal file
@@ -0,0 +1,199 @@
|
||||
import { createError, readBody } from "h3";
|
||||
import { writeFile, readFile, mkdir, rm } from "node:fs/promises";
|
||||
import { existsSync } from "node:fs";
|
||||
import { join, dirname } from "node:path";
|
||||
import { tmpdir } from "node:os";
|
||||
import { randomUUID } from "node:crypto";
|
||||
import { execFile } from "node:child_process";
|
||||
import { openai, TTS_MODEL, normalizeATC } from "../../utils/openai";
|
||||
|
||||
// --- Radio-Filter aus deinem bestehenden Endpoint wiederverwenden ---
|
||||
function buildRadioFilter(level: number) {
|
||||
const L = Math.max(1, Math.min(5, Math.floor(level || 4)));
|
||||
const dropout = (period: number, dur: number) =>
|
||||
`volume=enable='lt(mod(t\\,${period})\\,${dur})':volume=0`;
|
||||
const band = (hp: number, lp: number, gain = 6) =>
|
||||
`highpass=f=${hp},lowpass=f=${lp},compand=attacks=0.02:decays=0.25:points=-80/-900|-70/-20|0/-10|20/-8:gain=${gain}`;
|
||||
const tail = `aecho=0.6:0.7:8:0.08,acompressor=threshold=0.6:ratio=6:attack=20:release=200`;
|
||||
|
||||
const mkCrushMix = (bits: number, mix = 0.25, inLabel = "pre", outLabel = "mix1") => [
|
||||
`[${inLabel}]asplit=2[clean][toCrush]`,
|
||||
`[toCrush]acrusher=bits=${bits}:mix=1[crushed]`,
|
||||
`[clean][crushed]amix=inputs=2:weights=1 ${mix}:duration=shortest[${outLabel}]`,
|
||||
];
|
||||
|
||||
const mkNoiseMix = (amp: number, inLabel = "mix1", outLabel = "mix2") => [
|
||||
`anoisesrc=color=white:amplitude=${amp}[ns]`,
|
||||
`[${inLabel}][ns]amix=inputs=2:weights=1 0.25:duration=shortest[${outLabel}]`,
|
||||
];
|
||||
|
||||
const mkDropTail = (inLabel: string, period: number | null, dur: number | null) => {
|
||||
const chain: string[] = [];
|
||||
if (period && dur) chain.push(`[${inLabel}]${dropout(period, dur)}[drop]`);
|
||||
const src = period && dur ? "drop" : inLabel;
|
||||
chain.push(`[${src}]${tail}[out]`);
|
||||
return chain;
|
||||
};
|
||||
|
||||
if (L === 5) {
|
||||
return [
|
||||
`[0:a]${band(300,3400)},volume=1.2[a]`,
|
||||
`anoisesrc=color=white:amplitude=0.02[ns]`,
|
||||
`[a][ns]amix=inputs=2:weights=1 0.25:duration=shortest[mix]`,
|
||||
`[mix]${tail}[out]`,
|
||||
].join(";");
|
||||
}
|
||||
|
||||
if (L === 4) {
|
||||
return [
|
||||
`[0:a]${band(320,3300)},volume=1.15[pre]`,
|
||||
...mkNoiseMix(0.03, "pre", "mix"),
|
||||
`[mix]${tail}[out]`,
|
||||
].join(";");
|
||||
}
|
||||
|
||||
if (L === 3) {
|
||||
return [
|
||||
`[0:a]${band(350,3200)},volume=1.1[pre]`,
|
||||
...mkCrushMix(12, 0.22, "pre", "mix1"),
|
||||
...mkNoiseMix(0.05, "mix1", "mix2"),
|
||||
...mkDropTail("mix2", 6, 0.06),
|
||||
].join(";");
|
||||
}
|
||||
|
||||
if (L === 2) {
|
||||
return [
|
||||
`[0:a]${band(400,3000)},volume=1.05[pre]`,
|
||||
...mkCrushMix(10, 0.32, "pre", "mix1"),
|
||||
`anoisesrc=color=white:amplitude=0.08[ns]`,
|
||||
`[mix1][ns]amix=inputs=2:weights=1 0.6:duration=shortest[mix2]`,
|
||||
...mkDropTail("mix2", 4.5, 0.12),
|
||||
].join(";");
|
||||
}
|
||||
|
||||
// L === 1
|
||||
return [
|
||||
`[0:a]${band(500,2600,5)},volume=1.0[pre]`,
|
||||
...mkCrushMix(8, 0.45, "pre", "mix1"),
|
||||
`anoisesrc=color=white:amplitude=0.12[ns]`,
|
||||
`[mix1][ns]amix=inputs=2:weights=1 0.8:duration=shortest[mix2]`,
|
||||
...mkDropTail("mix2", 3.5, 0.2),
|
||||
].join(";");
|
||||
}
|
||||
|
||||
async function sh(cmd: string, args: string[]) {
|
||||
return new Promise<void>((res, rej) =>
|
||||
execFile(cmd, args, (err, _o, stderr) => (err ? rej(new Error(stderr || String(err))) : res()))
|
||||
);
|
||||
}
|
||||
|
||||
async function applyRadioEffect(input: string, outputWav: string, level = 4) {
|
||||
const filter = buildRadioFilter(level);
|
||||
await sh("ffmpeg", ["-y", "-i", input, "-filter_complex", filter, "-map", "[out]", "-ar", "16000", "-ac", "1", outputWav]);
|
||||
}
|
||||
|
||||
function outDir() {
|
||||
// Persistenz: env > ./storage/atc > tmp
|
||||
const base =
|
||||
process.env.ATC_OUT_DIR?.trim() ||
|
||||
join(process.cwd(), "storage", "atc");
|
||||
return base;
|
||||
}
|
||||
|
||||
async function ensureDir(p: string) {
|
||||
if (!existsSync(p)) await mkdir(p, { recursive: true });
|
||||
}
|
||||
|
||||
export default defineEventHandler(async (event) => {
|
||||
// POST { text: string; level?: 1..5; voice?: string; tag?: string }
|
||||
const body = await readBody<{
|
||||
text?: string;
|
||||
level?: number;
|
||||
voice?: string;
|
||||
tag?: string;
|
||||
}>(event);
|
||||
|
||||
const raw = (body?.text || "").trim();
|
||||
if (!raw) throw createError({ statusCode: 400, statusMessage: "text required" });
|
||||
|
||||
const level = Math.max(1, Math.min(5, Math.floor(body?.level ?? 4)));
|
||||
const voice = (body?.voice || "alloy") as string;
|
||||
|
||||
// Wortstamm/Normalisierung: ICAO-Formatierung, Zahl-/Buchstabierlogik etc.
|
||||
const normalized = normalizeATC(raw);
|
||||
if (!normalized) throw createError({ statusCode: 400, statusMessage: "normalized text empty" });
|
||||
|
||||
// 1) TTS (clean) -> tmp WAV
|
||||
const tmpClean = join(tmpdir(), `tts-${randomUUID()}.wav`);
|
||||
const tmpRadio = join(tmpdir(), `radio-${randomUUID()}.wav`);
|
||||
|
||||
const tts = await openai.audio.speech.create({
|
||||
model: TTS_MODEL,
|
||||
voice,
|
||||
format: "wav",
|
||||
input: normalized,
|
||||
});
|
||||
await writeFile(tmpClean, Buffer.from(await tts.arrayBuffer()));
|
||||
|
||||
// 2) Radio-Effekt (WAV, 16 kHz mono)
|
||||
await applyRadioEffect(tmpClean, tmpRadio, level);
|
||||
|
||||
// 3) Platzsparend transkodieren: OGG/Opus (VoIP-Profil), ~12 kbps @16 kHz mono
|
||||
const id = randomUUID();
|
||||
const baseDir = join(outDir(), new Date().toISOString().slice(0, 10)); // YYYY-MM-DD
|
||||
const fileOgg = join(baseDir, `${id}.ogg`);
|
||||
const fileJson = join(baseDir, `${id}.json`);
|
||||
|
||||
await ensureDir(baseDir);
|
||||
await sh("ffmpeg", [
|
||||
"-y",
|
||||
"-i", tmpRadio,
|
||||
"-ac", "1",
|
||||
"-ar", "16000",
|
||||
"-c:a", "libopus",
|
||||
"-b:a", "12k",
|
||||
"-application", "voip",
|
||||
fileOgg,
|
||||
]);
|
||||
|
||||
// 4) JSON-Metadaten neben die Audiodatei schreiben
|
||||
const meta = {
|
||||
id,
|
||||
createdAt: new Date().toISOString(),
|
||||
level,
|
||||
voice,
|
||||
text: raw,
|
||||
normalized,
|
||||
tag: body?.tag || null,
|
||||
files: {
|
||||
ogg: fileOgg,
|
||||
},
|
||||
model: TTS_MODEL,
|
||||
format: "audio/ogg; codecs=opus",
|
||||
};
|
||||
await writeFile(fileJson, JSON.stringify(meta, null, 2), "utf-8");
|
||||
|
||||
// 5) Response (Audio als Base64 + Meta). Optional: nur Pfad zurückgeben, wenn du sparen willst.
|
||||
const oggB64 = (await readFile(fileOgg)).toString("base64");
|
||||
|
||||
// Cleanup tmp
|
||||
rm(tmpClean).catch(() => {});
|
||||
rm(tmpRadio).catch(() => {});
|
||||
|
||||
return {
|
||||
ok: true,
|
||||
id,
|
||||
level,
|
||||
voice,
|
||||
text: raw,
|
||||
normalized,
|
||||
audio: {
|
||||
mime: "audio/ogg",
|
||||
base64: oggB64,
|
||||
},
|
||||
stored: {
|
||||
audioPath: fileOgg,
|
||||
jsonPath: fileJson,
|
||||
},
|
||||
};
|
||||
});
|
||||
154
server/api/atc/say.stream.get.ts
Normal file
154
server/api/atc/say.stream.get.ts
Normal file
@@ -0,0 +1,154 @@
|
||||
import { getQuery, createError } from "h3";
|
||||
import { spawn } from "node:child_process";
|
||||
import { openai, TTS_MODEL, normalizeATC } from "../../utils/openai";
|
||||
|
||||
function buildRadioFilter(level: number) {
|
||||
const L = Math.max(1, Math.min(5, Math.floor(level || 4)));
|
||||
const band = (hp: number, lp: number, gain = 6) =>
|
||||
`highpass=f=${hp},lowpass=f=${lp},compand=attacks=0.02:decays=0.25:points=-80/-900|-70/-20|0/-10|20/-8:gain=${gain}`;
|
||||
const dropout = (period: number, dur: number) =>
|
||||
`volume=enable='lt(mod(t\\,${period})\\,${dur})':volume=0`;
|
||||
const tail = `aecho=0.6:0.7:8:0.08,acompressor=threshold=0.6:ratio=6:attack=20:release=200`;
|
||||
|
||||
// helpers (ACHTUNG: weights **gequotet** oder mit | getrennt)
|
||||
const mixCrush = (bits: number, mix = 0.25, inLabel = "pre", outLabel = "mix1") => [
|
||||
`[${inLabel}]asplit=2[clean][toCrush]`,
|
||||
`[toCrush]acrusher=bits=${bits}:mix=1[crushed]`,
|
||||
`[clean][crushed]amix=inputs=2:weights='1 ${mix}':duration=shortest[${outLabel}]`,
|
||||
];
|
||||
|
||||
const mixNoise = (amp: number, inLabel = "mix1", outLabel = "mix2") => [
|
||||
`anoisesrc=color=white:amplitude=${amp}[ns]`,
|
||||
`[${inLabel}][ns]amix=inputs=2:weights='1 0.25':duration=shortest[${outLabel}]`,
|
||||
];
|
||||
|
||||
const dropAndTail = (inLabel: string, period?: number, dur?: number) => {
|
||||
const chain: string[] = [];
|
||||
if (period && dur) chain.push(`[${inLabel}]${dropout(period, dur)}[drop]`);
|
||||
const src = period && dur ? "drop" : inLabel;
|
||||
chain.push(`[${src}]${tail}[out]`);
|
||||
return chain;
|
||||
};
|
||||
|
||||
if (L === 5)
|
||||
return [
|
||||
`[0:a]${band(300,3400)},volume=1.2[a]`,
|
||||
`anoisesrc=color=white:amplitude=0.02[ns]`,
|
||||
`[a][ns]amix=inputs=2:weights='1 0.25':duration=shortest[mix]`,
|
||||
`[mix]${tail}[out]`,
|
||||
].join(";");
|
||||
|
||||
if (L === 4)
|
||||
return [
|
||||
`[0:a]${band(320,3300)},volume=1.15[pre]`,
|
||||
...mixNoise(0.03, "pre", "mix"),
|
||||
`[mix]${tail}[out]`,
|
||||
].join(";");
|
||||
|
||||
if (L === 3)
|
||||
return [
|
||||
`[0:a]${band(350,3200)},volume=1.1[pre]`,
|
||||
...mixCrush(12, 0.22, "pre", "mix1"),
|
||||
...mixNoise(0.05, "mix1", "mix2"),
|
||||
...dropAndTail("mix2", 6, 0.06),
|
||||
].join(";");
|
||||
|
||||
if (L === 2)
|
||||
return [
|
||||
`[0:a]${band(400,3000)},volume=1.05[pre]`,
|
||||
...mixCrush(10, 0.32, "pre", "mix1"),
|
||||
`anoisesrc=color=white:amplitude=0.08[ns]`,
|
||||
`[mix1][ns]amix=inputs=2:weights='1 0.6':duration=shortest[mix2]`,
|
||||
...dropAndTail("mix2", 4.5, 0.12),
|
||||
].join(";");
|
||||
|
||||
// L === 1
|
||||
return [
|
||||
`[0:a]${band(500,2600,5)},volume=1.0[pre]`,
|
||||
...mixCrush(8, 0.45, "pre", "mix1"),
|
||||
`anoisesrc=color=white:amplitude=0.12[ns]`,
|
||||
`[mix1][ns]amix=inputs=2:weights='1 0.8':duration=shortest[mix2]`,
|
||||
...dropAndTail("mix2", 3.5, 0.2),
|
||||
].join(";");
|
||||
}
|
||||
|
||||
// super-simpler Fallback ohne Mix/Noise (falls Filter scheitert)
|
||||
const SIMPLE_FILTER = `[0:a]highpass=f=350,lowpass=f=3000,acompressor=threshold=0.6:ratio=6:attack=20:release=200[out]`;
|
||||
|
||||
export default defineEventHandler(async (event) => {
|
||||
const q = getQuery(event);
|
||||
const raw = String(q.text || "").trim();
|
||||
if (!raw) throw createError({ statusCode: 400, statusMessage: "text required" });
|
||||
|
||||
const level = Math.max(1, Math.min(5, parseInt(String(q.level ?? "4"), 10) || 4));
|
||||
const voice = String(q.voice || "alloy");
|
||||
const normalized = normalizeATC(raw) || raw;
|
||||
|
||||
// 1) TTS → WAV
|
||||
const tts = await openai.audio.speech.create({
|
||||
model: TTS_MODEL,
|
||||
voice,
|
||||
format: "wav",
|
||||
input: normalized,
|
||||
});
|
||||
const wav = Buffer.from(await tts.arrayBuffer());
|
||||
if (wav.byteLength < 100) throw createError({ statusCode: 500, statusMessage: "TTS empty" });
|
||||
|
||||
async function runFfmpeg(filter: string) {
|
||||
return new Promise<void>((resolve) => {
|
||||
const ff = spawn("ffmpeg", [
|
||||
"-hide_banner", "-loglevel", "error",
|
||||
"-f", "wav", "-i", "pipe:0",
|
||||
"-filter_complex", filter,
|
||||
"-map", "[out]",
|
||||
"-ac", "1", "-ar", "16000",
|
||||
"-c:a", "libopus", "-b:a", "12k", "-application", "voip",
|
||||
"-f", "ogg", "pipe:1",
|
||||
], { stdio: ["pipe", "pipe", "pipe"] });
|
||||
|
||||
const res = event.node.res;
|
||||
let started = false;
|
||||
let ffErr = "";
|
||||
|
||||
ff.stderr.on("data", d => { ffErr += d.toString(); });
|
||||
|
||||
ff.stdout.once("data", (chunk: Buffer) => {
|
||||
if (!started) {
|
||||
started = true;
|
||||
res.statusCode = 200;
|
||||
res.setHeader("Content-Type", "audio/ogg");
|
||||
res.setHeader("Cache-Control", "no-store");
|
||||
res.setHeader("Accept-Ranges", "none");
|
||||
(res as any).flushHeaders?.();
|
||||
}
|
||||
res.write(chunk);
|
||||
ff.stdout.pipe(res);
|
||||
});
|
||||
|
||||
ff.stdin.on("error", () => {});
|
||||
ff.stdin.write(wav);
|
||||
ff.stdin.end();
|
||||
|
||||
ff.on("close", (code) => {
|
||||
if (!started) {
|
||||
// Fehlerpfad → JSON-Fehler zurück
|
||||
if (!res.headersSent) {
|
||||
res.statusCode = 500;
|
||||
res.setHeader("Content-Type", "application/json");
|
||||
}
|
||||
res.end(JSON.stringify({ error: true, message: ffErr || `ffmpeg exit ${code}` }));
|
||||
} else {
|
||||
if (!res.writableEnded) res.end();
|
||||
}
|
||||
resolve();
|
||||
});
|
||||
});
|
||||
}
|
||||
|
||||
// erst komplexer Filter; wenn der fehlschlägt → SIMPLE_FILTER
|
||||
await runFfmpeg(buildRadioFilter(level));
|
||||
if (!event.node.res.headersSent) {
|
||||
// zweiter Versuch
|
||||
await runFfmpeg(SIMPLE_FILTER);
|
||||
}
|
||||
});
|
||||
Reference in New Issue
Block a user