feat: switch TTS to edge-tts en-US-GuyNeural (neutral US accent)

Author: srpatcha

.github/workflows/video-build.yml

@@ -32,7 +32,7 @@ jobs:
         run: |
           sudo apt-get update
           sudo apt-get install -y ffmpeg libcairo2-dev libpango1.0-dev
-          pip install manim==0.18.1 gTTS mutagen
+          pip install manim==0.18.1 edge-tts mutagen
 
       - name: Generate narration audio (per-segment)
         run: |
@@ -73,3 +73,4 @@ jobs:
           fail_on_unmatched_files: false
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+

promo/generate_audio.py

@@ -1,45 +1,54 @@
-"""Generate per-segment narration audio and output durations for Manim sync."""
+"""Generate per-segment narration using edge-tts (US English neural voice)."""
+import asyncio
 import json
-from gtts import gTTS
+import edge_tts
 from mutagen.mp3 import MP3
 
+# en-US-GuyNeural = neutral male US voice (Silicon Valley style)
+VOICE = "en-US-GuyNeural"
+RATE = "+0%"  # natural pace
+
 SEGMENTS = [
-    {"id": "intro", "text": "Introducing ebuild. EoS Embedded Build System."},
+    {"id": "intro", "text": "Introducing ebuild. The embedded build system for EoS."},
     {"id": "f1", "text": "Feature one. Cross-Compilation Toolchains. Target ARM, RISC-V, and x86 from a single host with automatic toolchain management."},
     {"id": "f2", "text": "Feature two. Incremental Builds. Content-addressed caching rebuilds only changed targets, cutting build times by 90 percent."},
     {"id": "f3", "text": "Feature three. Dependency Resolution. DAG-based resolver handles diamond dependencies and circular includes automatically."},
-    {"id": "arch", "text": "Under the hood, ebuild is built with C, Make, CMake, LLVM. The architecture flows from Parser, to Resolver, to Scheduler, to Compiler, to Linker."},
-    {"id": "cta", "text": "ebuild. Open source and production ready. Visit github dot com slash embeddedos-org slash ebuild."},
+    {"id": "arch", "text": "Under the hood, ebuild is built with C, Make, CMake, and LLVM. The architecture flows from Parser, to Resolver, to Scheduler, to Compiler, to Linker."},
+    {"id": "cta", "text": "ebuild. Open source and production ready. Visit github dot com slash embeddedos-org slash ebuild."}
 ]
 
-durations = {}
-audio_files = []
-
-for seg in SEGMENTS:
-    filename = f"seg_{seg['id']}.mp3"
-    tts = gTTS(text=seg["text"], lang="en", slow=False)
-    tts.save(filename)
-    dur = MP3(filename).info.length
-    durations[seg["id"]] = round(dur + 0.5, 1)  # add 0.5s padding
-    audio_files.append(filename)
-    print(f"  {seg['id']}: {dur:.1f}s -> padded {durations[seg['id']]}s")
-
-# Write durations JSON for Manim to read
-with open("durations.json", "w") as f:
-    json.dump(durations, f, indent=2)
-
-# Concatenate all segments into single narration.mp3
-import subprocess
-list_file = "concat_list.txt"
-with open(list_file, "w") as f:
-    for af in audio_files:
-        f.write(f"file '{af}'\n")
-
-subprocess.run([
-    "ffmpeg", "-y", "-f", "concat", "-safe", "0",
-    "-i", list_file, "-c", "copy", "narration.mp3"
-], check=True)
-
-total = sum(durations.values())
-print(f"\nTotal narration: {total:.1f}s")
-print(f"Durations: {json.dumps(durations)}")
+
+async def generate():
+    durations = {}
+    audio_files = []
+
+    for seg in SEGMENTS:
+        filename = f"seg_{seg['id']}.mp3"
+        communicate = edge_tts.Communicate(seg["text"], VOICE, rate=RATE)
+        await communicate.save(filename)
+        dur = MP3(filename).info.length
+        durations[seg["id"]] = round(dur + 0.5, 1)
+        audio_files.append(filename)
+        print(f"  {seg['id']}: {dur:.1f}s -> padded {durations[seg['id']]}s")
+
+    with open("durations.json", "w") as f:
+        json.dump(durations, f, indent=2)
+
+    # Concatenate
+    import subprocess
+    with open("concat_list.txt", "w") as f:
+        for af in audio_files:
+            f.write(f"file '{af}'\n")
+
+    subprocess.run([
+        "ffmpeg", "-y", "-f", "concat", "-safe", "0",
+        "-i", "concat_list.txt", "-c", "copy", "narration.mp3"
+    ], check=True)
+
+    total = sum(durations.values())
+    print(f"\nVoice: {VOICE}")
+    print(f"Total narration: {total:.1f}s")
+    print(f"Durations: {json.dumps(durations)}")
+
+
+asyncio.run(generate())