feature: implement base tts with uv and kokoro

tts.ts

@@ -0,0 +1,40 @@
+import chalk from "chalk";
+
+const text = await Bun.stdin.text();
+
+if (!text) process.exit(1);
+
+const koko = Bun.spawn({
+  cmd: ["uv", "run", "--", "python", "-u", "main.py"],
+  cwd: "./tts/",
+  stdin: Buffer.from(text),
+  stdout: "pipe",
+  stderr: "inherit",
+});
+
+const decoder = new TextDecoder();
+for await (const line of koko.stdout) {
+  const decoded = decoder.decode(line).trim();
+
+  let infos: { file: number; text: string }[];
+  try {
+    infos = Bun.JSONL.parse(decoded) as { file: number; text: string }[];
+  } catch {
+    console.log(decoded);
+    continue;
+  }
+
+  for (const info of infos) {
+    console.log(`\n${chalk.blue(info.file)}: ${info.text}\n`);
+
+    await Bun.spawn({
+      cmd: ["mpv", `${info.file}.wav`],
+      cwd: "./tts/",
+      stdin: "ignore",
+      stdout: "inherit",
+      stderr: "inherit",
+    }).exited;
+  }
+}
+
+await koko.exited;

tts/main.py

@@ -0,0 +1,13 @@
+from kokoro import KPipeline
+import soundfile as sf
+import sys
+import json
+
+text = sys.stdin.read()
+
+pipeline = KPipeline(lang_code="a", repo_id="hexgrad/Kokoro-82M")
+generator = pipeline(text, voice="af_heart", speed=1.2)
+
+for i, (gs, ps, audio) in enumerate(generator):
+    print(json.dumps({"file": i, "text": gs}))
+    sf.write(f"{i}.wav", audio, 24000)

tts/pyproject.toml

@@ -0,0 +1,8 @@
+[project]
+name = "tts"
+version = "0.1.0"
+requires-python = ">=3.12"
+dependencies = [
+    "kokoro>=0.9.4",
+    "soundfile>=0.13.1",
+]