Add jingle generator for Radio Susan

1 files changed, 282 insertions, 0 deletions

diff --git a/generate_jingles.py b/generate_jingles.py
new file mode 100644
index 0000000..bdaa212
--- /dev/null
+++ b/generate_jingles.py
@@ -0,0 +1,282 @@
+#!/usr/bin/env python3
+"""
+Jingle Factory for Radio Susan
+
+Phrases use {sfx_name} markers to insert sound effects at breakpoints.
+Each phrase gets TTS'd with a random voice, then stitched with SFX.
+Background ambience (static/noise) is layered underneath.
+
+Usage: python3 generate_jingles.py [--list-sfx] [--preview PHRASE]
+
+SFX are stored in /var/lib/radio/sfx/
+Jingles output to /var/lib/radio/jingles/
+"""
+
+import asyncio
+import json
+import os
+import random
+import re
+import subprocess
+import sys
+import tempfile
+from pathlib import Path
+
+RADIO_DIR = Path("/var/lib/radio")
+SFX_DIR = RADIO_DIR / "sfx"
+JINGLES_DIR = RADIO_DIR / "jingles"
+VENV_BIN = RADIO_DIR / "venv" / "bin"
+EDGE_TTS = VENV_BIN / "edge-tts"
+
+# Voices to randomly pick from — mix of dramatic, cheesy, authoritative
+VOICES = [
+    "en-US-GuyNeural",          # Passion — classic radio voice
+    "en-US-ChristopherNeural",  # Authority — news anchor
+    "en-US-EricNeural",         # Rational — deadpan
+    "en-US-BrianNeural",        # Casual bro
+    "en-GB-RyanNeural",         # British bloke
+    "en-GB-ThomasNeural",       # Posh British
+    "en-US-AriaNeural",         # Confident woman
+    "en-US-JennyNeural",        # Friendly woman
+    "en-GB-SoniaNeural",        # British woman
+]
+
+# Phrase definitions — {sfx_name} marks where sound effects go
+# Add as many as you want!
+PHRASES = [
+    "You're listening to {crash} ninety nine point oh {airhorn} Susan Radio!",
+    "This is {static} Susan Radio {crash} ninety nine point oh {reverb_hit} all day, all night",
+    "{static} You're tuned in to {crash} Susan Radio! {airhorn}",
+    "Susan Radio {reverb_hit} ninety nine point oh {crash} on your dial",
+    "{static} Don't touch that dial! {crash} Susan Radio {airhorn} ninety nine point oh!",
+    "Coming at you live {crash} this is {reverb_hit} Susan Radio! {airhorn}",
+    "{static} Lock it in {crash} ninety nine point oh {reverb_hit} Susan Radio!",
+]
+
+
+def generate_sfx():
+    """Generate built-in sound effects using ffmpeg."""
+    SFX_DIR.mkdir(parents=True, exist_ok=True)
+    
+    sfx_specs = {
+        "crash": {
+            # White noise burst with sharp attack, fast decay
+            "filter": "anoisesrc=d=0.8:c=white:a=0.6,afade=t=in:d=0.02,afade=t=out:st=0.15:d=0.65,lowpass=f=3000",
+        },
+        "airhorn": {
+            # Stacked sine waves with vibrato for that MLG airhorn feel
+            "filter": (
+                "sine=f=750:d=0.6,aformat=sample_rates=44100"
+            ),
+            "filter_complex": (
+                "[0:a]volume=0.4[a];"
+                "anoisesrc=d=0.6:c=pink:a=0.1,aformat=sample_rates=44100[n];"
+                "sine=f=950:d=0.6,aformat=sample_rates=44100,volume=0.3[b];"
+                "[a][n]amix=inputs=2[m1];[m1][b]amix=inputs=2,afade=t=in:d=0.02,afade=t=out:st=0.3:d=0.3[out]"
+            ),
+        },
+        "static": {
+            # Pink noise crackle
+            "filter": "anoisesrc=d=0.5:c=pink:a=0.25,afade=t=in:d=0.1,afade=t=out:st=0.3:d=0.2,highpass=f=500",
+        },
+        "reverb_hit": {
+            # Impact with reverb tail
+            "filter": "anoisesrc=d=0.05:c=white:a=0.8,apad=whole_dur=1,aecho=0.8:0.88:60:0.4,afade=t=out:st=0.3:d=0.7",
+        },
+        "sweep": {
+            # Frequency sweep
+            "filter": "sine=f=200:d=0.4,aformat=sample_rates=44100,afade=t=in:d=0.05,afade=t=out:st=0.2:d=0.2,vibrato=f=20:d=1.0",
+        },
+    }
+    
+    for name, spec in sfx_specs.items():
+        outfile = SFX_DIR / f"{name}.wav"
+        if outfile.exists():
+            continue
+            
+        if "filter_complex" in spec:
+            cmd = [
+                "ffmpeg", "-y",
+                "-f", "lavfi", "-i", spec["filter"],
+                "-filter_complex", spec["filter_complex"],
+                "-map", "[out]",
+                "-ar", "44100", "-ac", "2",
+                str(outfile)
+            ]
+        else:
+            cmd = [
+                "ffmpeg", "-y",
+                "-f", "lavfi", "-i", spec["filter"],
+                "-ar", "44100", "-ac", "2",
+                str(outfile)
+            ]
+        
+        subprocess.run(cmd, capture_output=True)
+        print(f"  Generated SFX: {name}")
+    
+    return {f.stem: str(f) for f in SFX_DIR.glob("*.wav")}
+
+
+def tts_segment(text, voice, outfile):
+    """Generate TTS for a text segment using edge-tts."""
+    cmd = [
+        str(EDGE_TTS),
+        "--voice", voice,
+        "--rate", "+10%",  # Slightly faster for radio energy
+        "--text", text,
+        "--write-media", outfile,
+    ]
+    result = subprocess.run(cmd, capture_output=True, text=True)
+    if result.returncode != 0:
+        print(f"  TTS error: {result.stderr}")
+        return False
+    return True
+
+
+def stitch_jingle(segments, sfx_files, output_path):
+    """
+    Stitch TTS segments and SFX together with background ambience.
+    segments: list of (type, path) where type is 'voice' or 'sfx'
+    """
+    if not segments:
+        return False
+    
+    # Build ffmpeg concat filter
+    inputs = []
+    filter_parts = []
+    
+    for i, (stype, path) in enumerate(segments):
+        inputs.extend(["-i", path])
+        # Normalize all inputs to same format
+        filter_parts.append(f"[{i}:a]aformat=sample_rates=44100:channel_layouts=stereo,aresample=44100[s{i}]")
+    
+    # Concat all segments
+    concat_inputs = "".join(f"[s{i}]" for i in range(len(segments)))
+    filter_parts.append(f"{concat_inputs}concat=n={len(segments)}:v=0:a=1[main]")
+    
+    # Generate background static layer
+    filter_parts.append(
+        "anoisesrc=d=30:c=pink:a=0.04,aformat=sample_rates=44100:channel_layouts=stereo,"
+        "highpass=f=800,lowpass=f=4000[bg]"
+    )
+    
+    # Mix main audio with background, trim to main length
+    filter_parts.append(
+        "[bg][main]amix=inputs=2:duration=shortest:weights=1 3,"
+        "afade=t=in:d=0.1,volume=2.0[out]"
+    )
+    
+    filter_complex = ";\n".join(filter_parts)
+    
+    cmd = [
+        "ffmpeg", "-y",
+        *inputs,
+        "-filter_complex", filter_complex,
+        "-map", "[out]",
+        "-ar", "44100", "-ac", "2", "-b:a", "192k",
+        str(output_path)
+    ]
+    
+    result = subprocess.run(cmd, capture_output=True, text=True)
+    if result.returncode != 0:
+        print(f"  FFmpeg error: {result.stderr[-500:]}")
+        return False
+    return True
+
+
+def parse_phrase(phrase):
+    """Parse a phrase into segments: text parts and {sfx} markers."""
+    parts = re.split(r'\{(\w+)\}', phrase)
+    segments = []
+    for i, part in enumerate(parts):
+        part = part.strip()
+        if not part:
+            continue
+        if i % 2 == 0:
+            segments.append(("text", part))
+        else:
+            segments.append(("sfx", part))
+    return segments
+
+
+def generate_one_jingle(phrase, voice, sfx_files, output_path):
+    """Generate a single jingle from a phrase template."""
+    parsed = parse_phrase(phrase)
+    
+    audio_segments = []  # (type, filepath)
+    tmpfiles = []
+    
+    try:
+        for stype, content in parsed:
+            if stype == "text":
+                tmpfile = tempfile.mktemp(suffix=".mp3", dir="/tmp")
+                tmpfiles.append(tmpfile)
+                if tts_segment(content, voice, tmpfile):
+                    audio_segments.append(("voice", tmpfile))
+                else:
+                    print(f"  Failed TTS for: {content}")
+                    return False
+            elif stype == "sfx":
+                if content in sfx_files:
+                    audio_segments.append(("sfx", sfx_files[content]))
+                else:
+                    print(f"  Unknown SFX: {content}")
+        
+        if not audio_segments:
+            return False
+            
+        return stitch_jingle(audio_segments, sfx_files, output_path)
+    finally:
+        for f in tmpfiles:
+            try:
+                os.unlink(f)
+            except:
+                pass
+
+
+def main():
+    JINGLES_DIR.mkdir(parents=True, exist_ok=True)
+    
+    if "--list-sfx" in sys.argv:
+        sfx_files = generate_sfx()
+        print("Available SFX:", ", ".join(sorted(sfx_files.keys())))
+        return
+    
+    # Clear old jingles before regenerating
+    for old in JINGLES_DIR.glob("jingle_*.mp3"):
+        old.unlink()
+    
+    print("Generating SFX...")
+    sfx_files = generate_sfx()
+    print(f"  {len(sfx_files)} SFX ready: {', '.join(sorted(sfx_files.keys()))}")
+    
+    # Generate multiple jingles with different voice/phrase combos
+    print("\nGenerating jingles...")
+    
+    generated = 0
+    for i, phrase in enumerate(PHRASES):
+        voice = random.choice(VOICES)
+        output = JINGLES_DIR / f"jingle_{i+1:02d}.mp3"
+        
+        print(f"\n  [{i+1}/{len(PHRASES)}] Voice: {voice}")
+        print(f"  Phrase: {phrase}")
+        
+        if generate_one_jingle(phrase, voice, sfx_files, output):
+            # Get duration
+            probe = subprocess.run(
+                ["ffprobe", "-hide_banner", "-show_entries", "format=duration",
+                 "-of", "csv=p=0", str(output)],
+                capture_output=True, text=True
+            )
+            dur = float(probe.stdout.strip()) if probe.stdout.strip() else 0
+            print(f"  ✓ {output.name} ({dur:.1f}s)")
+            generated += 1
+        else:
+            print(f"  ✗ Failed")
+    
+    print(f"\nDone! {generated}/{len(PHRASES)} jingles generated in {JINGLES_DIR}")
+    print("Reload liquidsoap to pick them up: sudo systemctl restart liquidsoap")
+
+
+if __name__ == "__main__":
+    main()