takomi 2.1.41 → 2.1.43

package/assets/.agent/skills/hyperframes-media/scripts/lib/tts.mjs

@@ -0,0 +1,298 @@
+// tts.mjs — multi-provider TTS for the media audio engine. The provider chain,
+// auto-detected from env, is the one documented in ../SKILL.md:
+//
+//   1. HeyGen (Starfish)  — $HEYGEN_API_KEY / $HYPERFRAMES_API_KEY / ~/.heygen.
+//        Direct v3 REST (NOT `hyperframes tts`, which in the published build is
+//        Kokoro-only and silently ignores a HeyGen key). Returns word_timestamps
+//        in the same call, so no separate transcribe pass.
+//   2. ElevenLabs         — $ELEVENLABS_API_KEY + `pip install elevenlabs`. No
+//        word timings → caller chains transcribeWav().
+//   3. Kokoro-82M (local) — always available, via the published `hyperframes tts`
+//        CLI. No word timings → caller chains transcribeWav().
+//
+// "HeyGen available" is decided by CREDENTIAL presence (heygenCredential), never
+// by the CLI — see the note above.
+
+import { spawn, spawnSync } from "node:child_process";
+import { existsSync, mkdirSync, mkdtempSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { dirname, join } from "node:path";
+import { heygenAuthHeaders, heygenCredential, heygenJSON } from "./heygen.mjs";
+
+// ── provider detection ────────────────────────────────────────────────────────
+export function heygenAvailable() {
+  return heygenCredential() !== null;
+}
+export function elevenlabsAvailable() {
+  if (!process.env.ELEVENLABS_API_KEY) return false;
+  const r = spawnSync("python3", ["-c", "import elevenlabs"], {
+    stdio: "ignore",
+  });
+  return r.status === 0;
+}
+
+// First available provider wins; an explicit choice is honored (and validated).
+export function pickProvider(userProvider) {
+  if (userProvider) {
+    if (!["heygen", "elevenlabs", "kokoro"].includes(userProvider))
+      throw new Error(`invalid provider "${userProvider}" (heygen | elevenlabs | kokoro)`);
+    if (userProvider === "heygen" && !heygenAvailable())
+      throw new Error(
+        "provider=heygen but no HeyGen credentials (set $HEYGEN_API_KEY or run `npx hyperframes auth login`)",
+      );
+    if (userProvider === "elevenlabs" && !process.env.ELEVENLABS_API_KEY)
+      throw new Error("provider=elevenlabs but $ELEVENLABS_API_KEY is not set");
+    return userProvider;
+  }
+  return heygenAvailable() ? "heygen" : elevenlabsAvailable() ? "elevenlabs" : "kokoro";
+}
+
+// ── voice resolution ──────────────────────────────────────────────────────────
+// HeyGen /v3/voices/speech only accepts STARFISH voice_ids; auto-pick the first
+// English public starfish voice when none is pinned. ElevenLabs/Kokoro have
+// their own defaults.
+export async function resolveVoiceId({ provider, userVoice, lang = "en" }) {
+  if (userVoice) return userVoice;
+  if (provider === "elevenlabs") return "21m00Tcm4TlvDq8ikWAM"; // Rachel
+  if (provider === "kokoro") {
+    if (lang === "en") return "am_michael";
+    throw new Error("Kokoro non-English needs an explicit --voice (see references/tts.md)");
+  }
+  // heygen — pin a fixed English default so the choice is deterministic. The old
+  // "first English voice the API returns" drifts whenever HeyGen re-sorts the
+  // public catalog. Marcia (mature, low female). Override with --voice / request.voice.
+  if (lang === "en") return "05f19352e8f74b0392a8f411eba40de1"; // Marcia · English · female
+  // Non-English: no fixed default — fall back to the first matching catalog voice.
+  const payload = await heygenJSON(`/voices?engine=starfish&type=public&limit=50`, {
+    headers: heygenAuthHeaders(),
+  });
+  const voices = payload.data ?? payload.voices ?? [];
+  const pick = voices.find((v) => v.language === "English") ?? voices[0];
+  if (!pick) throw new Error("no public starfish voice to default to — pass --voice");
+  return pick.voice_id;
+}
+
+// ── helpers ─────────────────────────────────────────────────────────────────
+export function withWordIds(words) {
+  return (words ?? []).map((w, i) => ({
+    id: `w${i}`,
+    text: w.text,
+    start: w.start,
+    end: w.end,
+  }));
+}
+
+// `ffmpeg -i <file>` prints a `Duration: HH:MM:SS.ms` line to stderr even
+// though it exits non-zero with no output requested. Parsing pulled out as
+// a pure function so the ENOENT fallback below can be tested without
+// depending on whether ffprobe/ffmpeg are actually installed on the
+// machine running the tests.
+export function parseFfmpegDurationBanner(stderrText) {
+  const match = /Duration:\s*(\d+):(\d+):(\d+(?:\.\d+)?)/.exec(stderrText ?? "");
+  if (!match) return NaN;
+  const [, hours, minutes, seconds] = match;
+  return Number(hours) * 3600 + Number(minutes) * 60 + Number(seconds);
+}
+
+// Some "essentials"-style ffmpeg distributions (common on Windows) ship
+// ffmpeg.exe without ffprobe.exe. ffprobeDuration's caller (audio.mjs)
+// otherwise reads a spurious NaN as "the WAV file is corrupt" and drops an
+// already-successfully-synthesized TTS line, rather than "the tool for
+// measuring it is missing".
+function ffmpegDurationFallback(absPath) {
+  const r = spawnSync("ffmpeg", ["-i", absPath], { encoding: "utf8" });
+  return parseFfmpegDurationBanner(r.stderr);
+}
+
+export function ffprobeDuration(absPath) {
+  const r = spawnSync(
+    "ffprobe",
+    ["-v", "error", "-show_entries", "format=duration", "-of", "default=nw=1:nk=1", absPath],
+    { encoding: "utf8" },
+  );
+  if (r.error?.code === "ENOENT") return ffmpegDurationFallback(absPath);
+  if (r.status !== 0) return NaN;
+  return parseFloat(String(r.stdout).trim());
+}
+
+export function resolveNpxCliFromNpmExecPath(
+  npmExecPath = process.env.npm_execpath,
+  pathExists = existsSync,
+) {
+  if (!npmExecPath) return null;
+  const fileName = npmExecPath.replace(/\\/g, "/").split("/").pop()?.toLowerCase();
+  const npxCliPath =
+    fileName === "npx-cli.js" ? npmExecPath : join(dirname(npmExecPath), "npx-cli.js");
+  return pathExists(npxCliPath) ? npxCliPath : null;
+}
+
+export function resolveSpawnCommand(
+  cmd,
+  args,
+  opts = {},
+  platform = process.platform,
+  env = process.env,
+  pathExists = existsSync,
+) {
+  if (cmd !== "npx" || platform !== "win32") {
+    return { cmd, args, opts: { stdio: "ignore", ...opts } };
+  }
+
+  // On Windows, npx resolves to npx.cmd, which Node cannot execute directly.
+  // Avoid `shell:true` and the .cmd shim entirely by invoking npm's JS CLI with
+  // node, preserving request-provided values as argv data instead of shell text.
+  const npxCliPath = resolveNpxCliFromNpmExecPath(env.npm_execpath, pathExists);
+  if (!npxCliPath) return null;
+  return {
+    cmd: env.npm_node_execpath || process.execPath,
+    args: [npxCliPath, ...args.map((arg) => String(arg))],
+    opts: { stdio: "ignore", windowsHide: true, ...opts },
+  };
+}
+
+// `platform`/`spawnFn` params (default process.platform / the real spawn)
+// exist so tests can exercise the win32 branch without mocking node:child_process
+// (its ESM exports are non-configurable, so mock.method can't patch it).
+export function spawnP(
+  cmd,
+  args,
+  opts = {},
+  platform = process.platform,
+  spawnFn = spawn,
+  env = process.env,
+  pathExists = existsSync,
+) {
+  const resolved = resolveSpawnCommand(cmd, args, opts, platform, env, pathExists);
+  if (!resolved) return Promise.resolve({ status: -1 });
+  return new Promise((resolve) => {
+    const p = spawnFn(resolved.cmd, resolved.args, resolved.opts);
+    p.on("exit", (code) => resolve({ status: code ?? -1 }));
+    p.on("error", () => resolve({ status: -1 }));
+  });
+}
+
+// mp3/whatever bytes → wav 44.1k mono at destWav (ffmpeg detects true format).
+function transcodeToWav(bytes, destWav) {
+  const td = mkdtempSync(join(tmpdir(), "hf-tts-"));
+  const tmp = join(td, "a.mp3");
+  writeFileSync(tmp, bytes);
+  mkdirSync(dirname(destWav), { recursive: true });
+  const ff = spawnSync(
+    "ffmpeg",
+    ["-y", "-loglevel", "error", "-i", tmp, "-ar", "44100", "-ac", "1", destWav],
+    { stdio: "ignore" },
+  );
+  rmSync(td, { recursive: true, force: true });
+  return ff.status === 0 && existsSync(destWav);
+}
+
+const ELEVENLABS_PY = `
+import os, sys
+from elevenlabs.client import ElevenLabs
+from elevenlabs import save
+client = ElevenLabs(api_key=os.environ["ELEVENLABS_API_KEY"])
+text = open(sys.argv[1]).read()
+audio = client.text_to_speech.convert(
+    text=text, voice_id=sys.argv[2],
+    model_id="eleven_multilingual_v2", output_format="mp3_44100_128",
+)
+save(audio, sys.argv[3])
+`;
+
+// ── synthesize one line ───────────────────────────────────────────────────────
+// Writes wav at wavAbs. Returns { ok, words } — words is the raw
+// [{text,start,end}] array for HeyGen (native), or null for ElevenLabs/Kokoro
+// (caller must transcribeWav). Never throws; failures return { ok:false }.
+export async function synthesizeOne({
+  provider,
+  text,
+  voiceId,
+  lang = "en",
+  speed = 1.0,
+  wavAbs,
+  hyperframesDir,
+}) {
+  if (provider === "heygen") return synthesizeHeygen({ text, voiceId, lang, speed, wavAbs });
+  if (provider === "elevenlabs") {
+    const r = await spawnP(
+      "python3",
+      ["-c", ELEVENLABS_PY, writeTmpText(text), voiceId, wavAbs],
+      {},
+    );
+    return { ok: r.status === 0 && existsSync(wavAbs), words: null };
+  }
+  // kokoro — via the published CLI; --output is relative to the project dir.
+  const wavRel = relTo(hyperframesDir, wavAbs);
+  const args = ["hyperframes", "tts", writeTmpText(text), "--voice", voiceId, "--output", wavRel];
+  if (lang !== "en") args.push("--lang", lang);
+  const r = await spawnP("npx", args, { cwd: hyperframesDir });
+  return { ok: r.status === 0 && existsSync(wavAbs), words: null };
+}
+
+async function synthesizeHeygen({ text, voiceId, lang, speed, wavAbs }) {
+  try {
+    const body = { text, voice_id: voiceId, speed };
+    if (lang !== "en") body.language = lang;
+    const payload = await heygenJSON(`/voices/speech`, {
+      method: "POST",
+      headers: heygenAuthHeaders(),
+      body,
+    });
+    const inner = payload.data ?? payload;
+    if (!inner.audio_url) return { ok: false, words: null };
+    const res = await fetch(inner.audio_url);
+    if (!res.ok) return { ok: false, words: null };
+    const bytes = Buffer.from(await res.arrayBuffer());
+    // .wav output → transcode to 44.1k mono; .mp3 → raw bytes (no ffmpeg). The
+    // engine always asks for .wav; the standalone heygen-tts CLI may ask for .mp3.
+    if (wavAbs.endsWith(".wav")) {
+      if (!transcodeToWav(bytes, wavAbs)) return { ok: false, words: null };
+    } else {
+      mkdirSync(dirname(wavAbs), { recursive: true });
+      writeFileSync(wavAbs, bytes);
+    }
+    const words = Array.isArray(inner.word_timestamps)
+      ? inner.word_timestamps
+          .filter((w) => w && typeof w.word === "string" && isFinite(w.start) && isFinite(w.end))
+          .filter((w) => !/^<.*>$/.test(w.word.trim())) // drop <start>/<end> sentinels
+          .map((w) => ({ text: w.word, start: w.start, end: w.end }))
+      : [];
+    return { ok: true, words };
+  } catch {
+    return { ok: false, words: null };
+  }
+}
+
+// ElevenLabs/Kokoro have no word timings — run Whisper over the wav. Returns the
+// flat [{id,text,start,end}] word array, or null. Each call uses a throwaway
+// --dir so parallel scenes don't collide on transcript.json.
+export async function transcribeWav({ wavRel, lang = "en", hyperframesDir }) {
+  const model = lang === "en" ? "small.en" : "small";
+  const td = mkdtempSync(join(tmpdir(), "hf-trans-"));
+  const args = ["hyperframes", "transcribe", wavRel, "--model", model, "--dir", td];
+  if (lang !== "en") args.push("--language", lang);
+  const r = await spawnP("npx", args, { cwd: hyperframesDir });
+  let words = null;
+  if (r.status === 0) {
+    const src = join(td, "transcript.json");
+    if (existsSync(src)) {
+      try {
+        const arr = JSON.parse(readFileSync(src, "utf8"));
+        if (Array.isArray(arr) && arr.length) words = arr;
+      } catch {}
+    }
+  }
+  rmSync(td, { recursive: true, force: true });
+  return words;
+}
+
+// ── tiny local utils ──────────────────────────────────────────────────────────
+function writeTmpText(text) {
+  const td = mkdtempSync(join(tmpdir(), "hf-txt-"));
+  const p = join(td, "line.txt");
+  writeFileSync(p, text);
+  return p;
+}
+function relTo(base, abs) {
+  return abs.startsWith(base + "/") ? abs.slice(base.length + 1) : abs;
+}

package/assets/.agent/skills/hyperframes-media/scripts/lib/tts.spawn.test.mjs

@@ -0,0 +1,97 @@
+import { test } from "node:test";
+import assert from "node:assert/strict";
+import { EventEmitter } from "node:events";
+import { resolveNpxCliFromNpmExecPath, resolveSpawnCommand, spawnP } from "./tts.mjs";
+
+// Regression: on Windows, npx resolves to npx.cmd, which spawn() cannot exec
+// without shell:true — it fails ENOENT, silently swallowed as ok:false by the
+// caller. spawnP takes injectable platform/spawnFn params so this doesn't
+// need to touch the real process.platform or mock node:child_process (whose
+// ESM exports are non-configurable).
+function fakeSpawn(captured) {
+  return (cmd, args, opts) => {
+    captured.push({ cmd, args, opts });
+    const p = new EventEmitter();
+    setImmediate(() => p.emit("exit", 0));
+    return p;
+  };
+}
+
+const envWithNpxCli = {
+  npm_execpath: "/opt/node/lib/node_modules/npm/bin/npm-cli.js",
+  npm_node_execpath: "/opt/node/bin/node",
+};
+const npxCliPath = "/opt/node/lib/node_modules/npm/bin/npx-cli.js";
+const pathExists = (path) => path === npxCliPath;
+
+test("resolveNpxCliFromNpmExecPath finds npx-cli next to npm-cli", () => {
+  assert.equal(resolveNpxCliFromNpmExecPath(envWithNpxCli.npm_execpath, pathExists), npxCliPath);
+});
+
+test("resolveSpawnCommand routes npx through node+npx-cli on win32 without shell:true", () => {
+  const resolved = resolveSpawnCommand(
+    "npx",
+    ["hyperframes", "tts", "C:\\Users\\Test User\\line.txt", "--voice", "am_michael"],
+    {},
+    "win32",
+    envWithNpxCli,
+    pathExists,
+  );
+  assert.ok(resolved);
+  assert.equal(resolved.cmd, envWithNpxCli.npm_node_execpath);
+  assert.deepEqual(resolved.args, [
+    npxCliPath,
+    "hyperframes",
+    "tts",
+    "C:\\Users\\Test User\\line.txt",
+    "--voice",
+    "am_michael",
+  ]);
+  assert.equal(resolved.opts.shell, undefined);
+});
+
+test("resolveSpawnCommand preserves Windows npx shell metacharacters as argv data", () => {
+  const resolved = resolveSpawnCommand(
+    "npx",
+    ["hyperframes", "tts", "hello & calc"],
+    {},
+    "win32",
+    envWithNpxCli,
+    pathExists,
+  );
+  assert.ok(resolved);
+  assert.deepEqual(resolved.args, [npxCliPath, "hyperframes", "tts", "hello & calc"]);
+});
+
+test("spawnP uses the resolved node+npx-cli command for npx on win32", async () => {
+  const captured = [];
+  await spawnP(
+    "npx",
+    ["hyperframes", "tts"],
+    {},
+    "win32",
+    fakeSpawn(captured),
+    envWithNpxCli,
+    pathExists,
+  );
+  assert.equal(captured.length, 1);
+  assert.equal(captured[0].cmd, envWithNpxCli.npm_node_execpath);
+  assert.deepEqual(captured[0].args, [npxCliPath, "hyperframes", "tts"]);
+  assert.equal(captured[0].opts.shell, undefined);
+});
+
+test("spawnP does not enable shell for npx on darwin/linux", async () => {
+  const captured = [];
+  await spawnP("npx", ["hyperframes", "tts"], {}, "darwin", fakeSpawn(captured));
+  assert.equal(captured[0].cmd, "npx");
+  assert.deepEqual(captured[0].args, ["hyperframes", "tts"]);
+  assert.equal(captured[0].opts.shell, undefined);
+});
+
+test("spawnP does not enable shell for non-npx commands even on win32", async () => {
+  const captured = [];
+  await spawnP("python3", ["-c", "pass"], {}, "win32", fakeSpawn(captured));
+  assert.equal(captured[0].cmd, "python3");
+  assert.deepEqual(captured[0].args, ["-c", "pass"]);
+  assert.equal(captured[0].opts.shell, undefined);
+});

package/assets/.agent/skills/hyperframes-media/scripts/lib/tts.test.mjs

@@ -0,0 +1,66 @@
+import { test } from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, writeFileSync, chmodSync, rmSync } from "node:fs";
+import { join } from "node:path";
+import { tmpdir } from "node:os";
+import { parseFfmpegDurationBanner, ffprobeDuration } from "./tts.mjs";
+
+test("parseFfmpegDurationBanner reads ffmpeg's stderr Duration line", () => {
+  const stderr = [
+    "ffmpeg version 6.0",
+    "Input #0, wav, from 'a.wav':",
+    "  Duration: 00:00:03.42, bitrate: 705 kb/s",
+    "At least one output file must be specified",
+  ].join("\n");
+  assert.equal(parseFfmpegDurationBanner(stderr), 3.42);
+});
+
+test("parseFfmpegDurationBanner handles an hours component", () => {
+  const stderr = "  Duration: 01:02:03.50, start: 0.000000, bitrate: 128 kb/s";
+  assert.equal(parseFfmpegDurationBanner(stderr), 3723.5);
+});
+
+test("parseFfmpegDurationBanner returns NaN when there is no Duration line", () => {
+  assert.ok(Number.isNaN(parseFfmpegDurationBanner("ffmpeg: command not found")));
+  assert.ok(Number.isNaN(parseFfmpegDurationBanner("")));
+  assert.ok(Number.isNaN(parseFfmpegDurationBanner(undefined)));
+});
+
+// Regression for the actual bug: ffprobeDuration used to collapse "ffprobe
+// binary is missing" (ENOENT — the "essentials"-style Windows ffmpeg build
+// with no ffprobe.exe) and "file is genuinely unreadable" into the same NaN,
+// giving audio.mjs no way to tell "measure differently" from "give up".
+//
+// Builds an isolated PATH containing only a fake `ffmpeg` stub (no `ffprobe`
+// at all) so ffprobeDuration's spawnSync("ffprobe", ...) call ENOENTs for
+// real, then verifies it recovers the duration via the ffmpeg fallback
+// instead of returning NaN.
+test("ffprobeDuration falls back to ffmpeg when the ffprobe binary itself is missing", () => {
+  const dir = mkdtempSync(join(tmpdir(), "tts-ffprobe-fallback-"));
+  const fakeFfmpeg = join(dir, "ffmpeg");
+  writeFileSync(
+    fakeFfmpeg,
+    "#!/bin/sh\necho 'Duration: 00:00:02.50, start: 0.000000, bitrate: 128 kb/s' 1>&2\nexit 1\n",
+  );
+  chmodSync(fakeFfmpeg, 0o755);
+  const originalPath = process.env.PATH;
+  try {
+    process.env.PATH = dir; // only the fake ffmpeg resolves; no real ffprobe on this PATH
+    assert.equal(ffprobeDuration("/does/not/matter.wav"), 2.5);
+  } finally {
+    process.env.PATH = originalPath;
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("ffprobeDuration returns NaN when neither ffprobe nor ffmpeg resolve", () => {
+  const dir = mkdtempSync(join(tmpdir(), "tts-no-binaries-"));
+  const originalPath = process.env.PATH;
+  try {
+    process.env.PATH = dir; // empty directory — nothing resolves
+    assert.ok(Number.isNaN(ffprobeDuration("/does/not/matter.wav")));
+  } finally {
+    process.env.PATH = originalPath;
+    rmSync(dir, { recursive: true, force: true });
+  }
+});

package/assets/.agent/skills/hyperframes-media/scripts/lyria-recipe.py

@@ -0,0 +1,128 @@
+#!/usr/bin/env python3
+"""Generate BGM using Google Lyria RealTime API.
+
+Usage:
+    python lyria-recipe.py --output <path> --duration <seconds> [tuning flags]
+
+Requires:
+    $GOOGLE_API_KEY or $GEMINI_API_KEY environment variable (treated as aliases).
+    pip install google-genai python-dotenv. audio.mjs Step 4b installs these on
+    demand when a key is set but google.genai is not importable; if that install
+    fails it falls back to local MusicGen rather than leaving the video with no BGM.
+"""
+
+from __future__ import annotations
+
+import argparse
+import asyncio
+import os
+import sys
+import wave
+from pathlib import Path
+
+DEFAULT_PROMPT = "Uplifting corporate tech, bright and modern, gentle piano with synth pads"
+SAMPLE_RATE = 48000
+CHANNELS = 2
+SAMPLE_WIDTH = 2  # 16-bit
+
+
+def parse_args() -> argparse.Namespace:
+    p = argparse.ArgumentParser(description="Generate BGM via Google Lyria RealTime.")
+    p.add_argument("--output", required=True, help="Output WAV path.")
+    p.add_argument("--duration", type=float, required=True, help="Target duration in seconds.")
+    p.add_argument("--prompt", default=DEFAULT_PROMPT, help="Mood / instrumentation prompt.")
+    p.add_argument("--negative-prompt", default=None, help="Styles to exclude (optional).")
+    p.add_argument("--bpm", type=int, default=110)
+    p.add_argument("--brightness", type=float, default=0.8, help="0-1, higher = brighter mood.")
+    p.add_argument("--density", type=float, default=0.5, help="0-1, higher = fuller mix.")
+    p.add_argument(
+        "--scale",
+        default="MAJOR",
+        help="MAJOR / MINOR / PENTATONIC / etc. — see google.genai.types.Scale. Pass empty string for none.",
+    )
+    return p.parse_args()
+
+
+async def generate_bgm(args: argparse.Namespace) -> dict:
+    from google import genai
+    from google.genai import types
+
+    api_key = os.environ.get("GOOGLE_API_KEY") or os.environ.get("GEMINI_API_KEY") or ""
+    if not api_key:
+        raise RuntimeError("Neither GOOGLE_API_KEY nor GEMINI_API_KEY is set.")
+
+    client = genai.Client(
+        api_key=api_key,
+        http_options={"api_version": "v1alpha"},
+    )
+
+    out_path = Path(args.output)
+    out_path.parent.mkdir(parents=True, exist_ok=True)
+
+    target_bytes = int(args.duration * SAMPLE_RATE * CHANNELS * SAMPLE_WIDTH)
+
+    cfg: dict = {"bpm": args.bpm, "temperature": 1.0}
+    if args.density is not None:
+        cfg["density"] = args.density
+    if args.brightness is not None:
+        cfg["brightness"] = args.brightness
+    if args.scale:
+        scale_enum = getattr(types.Scale, args.scale, None)
+        if scale_enum:
+            cfg["scale"] = scale_enum
+
+    prompts = [types.WeightedPrompt(text=args.prompt, weight=1.0)]
+    if args.negative_prompt:
+        prompts.append(types.WeightedPrompt(text=args.negative_prompt, weight=-1.0))
+
+    buf = bytearray()
+    timeout = args.duration + 8
+
+    async with client.aio.live.music.connect(
+        model="models/lyria-realtime-exp",
+    ) as session:
+        await session.set_weighted_prompts(prompts=prompts)
+        await session.set_music_generation_config(
+            config=types.LiveMusicGenerationConfig(**cfg),
+        )
+        await session.play()
+
+        async def collect():
+            while len(buf) < target_bytes:
+                async for msg in session.receive():
+                    sc = msg.server_content
+                    if sc and sc.audio_chunks:
+                        for chunk in sc.audio_chunks:
+                            buf.extend(chunk.data)
+                            if len(buf) >= target_bytes:
+                                return
+                await asyncio.sleep(1e-6)
+
+        try:
+            await asyncio.wait_for(collect(), timeout=timeout)
+        except TimeoutError:
+            print(f"Timeout after {timeout:.0f}s, collected {len(buf)} bytes", file=sys.stderr)
+
+    audio = bytes(buf[:target_bytes])
+    with wave.open(str(out_path), "wb") as wf:
+        wf.setnchannels(CHANNELS)
+        wf.setsampwidth(SAMPLE_WIDTH)
+        wf.setframerate(SAMPLE_RATE)
+        wf.writeframes(audio)
+
+    actual_duration = len(audio) / (SAMPLE_RATE * CHANNELS * SAMPLE_WIDTH)
+    print(f"BGM: {out_path} ({actual_duration:.2f}s)")
+    return {"file": str(out_path), "duration_sec": round(actual_duration, 2)}
+
+
+def main() -> None:
+    args = parse_args()
+    try:
+        asyncio.run(generate_bgm(args))
+    except RuntimeError as exc:
+        print(f"BGM generation failed: {exc}", file=sys.stderr)
+        sys.exit(1)
+
+
+if __name__ == "__main__":
+    main()