Initial commit

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

saved/pokedex_voice_spike_scripts/README.md

@@ -0,0 +1,57 @@
+# Pokedex voice spike — scripts
+
+De-risking spike for the Pokédex app's robotic voice output. Full pipeline:
+text -> TTS -> optional DSP "robot" filter -> mp3.
+
+## Files
+
+- `generate_samples.sh` — end-to-end driver: installs are documented at the
+  top, then runs espeak-ng and Piper TTS, then the DSP filters, then
+  converts everything to mp3. Run it from this directory (`./generate_samples.sh`).
+- `robot_filter.py` — heavy/original robot filter: square-wave ring
+  modulation, bitcrush (sample-and-hold + bit-depth reduction), narrow
+  band-pass, hard drive. This is what "too robotic" was generated with.
+- `robot_filter_light.py` — barely-there version: sine-wave ring mod at
+  low mix, wide band-pass, mild soft-clip. Meant to add a hint of
+  synthesized edge to an already-natural neural voice without disturbing
+  cadence.
+- `robot_filter_v2.py` — the current one: parameterized, takes an
+  `intensity` argument from 0.0 (untouched) to 1.0 (full heavy filter) and
+  interpolates ring-mod mix/carrier, bandpass width, bitcrush, and drive
+  along that scale. Usage: `python3 robot_filter_v2.py in.wav out.wav 0.45`.
+  This is the one to keep tuning going forward — texted intensity numbers
+  ("try 0.6") map directly to its third argument.
+
+## Voice model
+
+TTS engine is [Piper](https://github.com/OHF-Voice/piper1-gpl) — a small,
+fully offline neural TTS with real Android ports available, which matters
+for the app's "works with no signal" requirement.
+
+The voice used is `en_US-joe-medium.onnx`, fine-tuned from Piper's stock
+"lessac" voice on a CC0-licensed dataset (see MODEL_CARD if you unpack the
+wheel) — no attribution/licensing issue to ship it.
+
+**How I got the model file in this sandbox:** the sandbox's network
+allowlist didn't reach huggingface.co, where Piper's official voices are
+hosted, only package registries. I found a community-published PyPI wheel
+(`joe-us-piper-voice`) that bundles the .onnx file directly, so a plain
+`pip install` pulled it down. That's a workaround specific to *this*
+sandbox — for the real project, get voices the normal way from Piper's
+official releases/voice list, which gives you far more voice choices
+(different speakers, accents, quality tiers) than this one bundled option.
+
+## What's still open
+
+- Pronunciation of individual Pokémon species names hasn't been
+  systematically checked — only "Pokémon," "Bulbasaur," "Charizard,"
+  "Pikachu" were tested. Expect to need a pass listening to all ~1000+
+  species names and hand-fixing the ones the phonemizer (espeak-ng, under
+  Piper's hood) gets wrong, either by respelling in the source text or via
+  espeak's phoneme-override escape syntax.
+- Robot filter intensity (`robot_filter_v2.py`'s `intensity` arg) needs to
+  land wherever the actual desired "amount of robotic" ends up — currently
+  parked at 0.45 pending feedback.
+- This whole pipeline is meant to run once, offline, over every dex entry
+  ahead of time (batch pre-generation), not live on-device — see the
+  earlier discussion in this conversation for why.

saved/pokedex_voice_spike_scripts/generate_all.py

@@ -0,0 +1,77 @@
+"""
+Windows-friendly regeneration of the voice spike, extended to all 5
+Pokemon the app currently knows about. Ports generate_samples.sh's
+"fixed" (cadence-corrected) + robot_filter_v2.py (intensity 0.45) stages
+to Python, using piper's Python API directly instead of the `piper` CLI
+and skipping the ffmpeg mp3 step (WAV plays fine on Android, and ffmpeg
+isn't installed on this machine).
+
+Usage: .venv\\Scripts\\python.exe generate_all.py
+"""
+
+import wave
+from pathlib import Path
+
+from piper import PiperVoice
+from piper.config import SynthesisConfig
+
+import robot_filter_v2
+
+HERE = Path(__file__).parent
+VOICE_MODEL = HERE / "voices" / "en_US-joe-medium.onnx"
+OUT_DIR = HERE / "output"
+
+ENTRIES = {
+    "bulbasaur": (
+        "Bulbasaur, the seed Pokémon. It can be seen napping in bright "
+        "sunlight. There is a seed on its back. By soaking up the sun's "
+        "rays, the seed grows progressively larger."
+    ),
+    "charizard": (
+        "Charizard, the flame Pokémon. Charizard flies around the sky in "
+        "search of powerful opponents. It breathes fire of such great "
+        "heat that it melts anything."
+    ),
+    "pikachu": (
+        "Pikachu, the mouse Pokémon. When several of these Pokémon "
+        "gather, their electricity could build and cause lightning "
+        "storms."
+    ),
+    "eevee": (
+        "Eevee, the evolution Pokémon. Its genetic code is irregular. It "
+        "may mutate if it is exposed to radiation from element stones."
+    ),
+    "squirtle": (
+        "Squirtle, the tiny turtle Pokémon. After birth, its back swells "
+        "and hardens into a shell. It powerfully sprays foam from its "
+        "mouth."
+    ),
+}
+
+ROBOT_INTENSITY = 0.45
+
+
+SYN_CONFIG = SynthesisConfig(noise_scale=0.5, noise_w_scale=0.3, length_scale=0.98)
+
+
+def main():
+    OUT_DIR.mkdir(exist_ok=True)
+    print(f"Loading voice model {VOICE_MODEL.name}...")
+    voice = PiperVoice.load(str(VOICE_MODEL))
+
+    for name, text in ENTRIES.items():
+        fixed_path = OUT_DIR / f"{name}_fixed.wav"
+        robot_path = OUT_DIR / f"{name}_fixed_robot.wav"
+        print(f"Synthesizing {name}...")
+        with wave.open(str(fixed_path), "wb") as wav_file:
+            voice.synthesize_wav(text, wav_file, syn_config=SYN_CONFIG)
+        sr, x = robot_filter_v2.load(str(fixed_path))
+        y = robot_filter_v2.robotize(x, sr, intensity=ROBOT_INTENSITY)
+        robot_filter_v2.save(str(robot_path), sr, y)
+        print(f"  -> {robot_path.name}")
+
+    print("\nDone. See output/*_fixed_robot.wav")
+
+
+if __name__ == "__main__":
+    main()

saved/pokedex_voice_spike_scripts/generate_samples.sh

@@ -0,0 +1,74 @@
+#!/usr/bin/env bash
+# Reproduces the Pokedex voice spike end to end:
+#   TTS (espeak-ng and Piper) -> robot-voice DSP filter -> mp3
+#
+# Setup (Debian/Ubuntu):
+#   sudo apt-get install -y espeak-ng ffmpeg
+#   pip install piper-tts --break-system-packages
+#   pip install numpy scipy --break-system-packages
+#
+# Voice model:
+#   Official route: download a Piper voice (.onnx + .onnx.json) from the
+#   Piper voices repo and point --model at it, e.g. en_US-lessac-medium.
+#   In this sandbox, huggingface.co wasn't reachable, so I instead pulled a
+#   community PyPI wheel that bundles the model file directly:
+#     pip download --no-deps joe-us-piper-voice
+#     (unzip the wheel; the .onnx/.onnx.json live under
+#      joe_us_piper_voice/data/). Not an official source -- for production,
+#      get voices from Piper's own releases instead.
+set -euo pipefail
+cd "$(dirname "$0")"
+
+VOICE_MODEL="voices/en_US-joe-medium.onnx"
+export ALSA_CONFIG_PATH=/dev/null   # silence ALSA warnings in headless envs
+
+# ---- Stage 1: raw espeak-ng (formant synth, offline, inherently "robotic") ----
+espeak-ng -v en-us -s 150 -w bulbasaur_raw.wav \
+  "Bulbasaur. Seed pokemon. It can be seen napping in bright sunlight. There is a seed on its back. By soaking up the sun's rays, the seed grows progressively larger."
+
+espeak-ng -v en-us -s 150 -w charizard_raw.wav \
+  "Charizard. Flame pokemon. Charizard flies around the sky in search of powerful opponents. It breathes fire of such great heat that it melts anything."
+
+espeak-ng -v en-us -s 150 -w pikachu_raw.wav \
+  "Pikachu. Mouse pokemon. When several of these pokemon gather, their electricity could build and cause lightning storms."
+
+# ---- Stage 2: Piper neural TTS, first pass (natural but bad cadence on rare words) ----
+gen_neural() {
+  local name="$1" text="$2"
+  echo "$text" | piper -m "$VOICE_MODEL" -f "${name}_neural.wav"
+}
+gen_neural bulbasaur "Bulbasaur. Seed pokemon. It can be seen napping in bright sunlight. There is a seed on its back. By soaking up the sun's rays, the seed grows progressively larger."
+gen_neural charizard "Charizard. Flame pokemon. Charizard flies around the sky in search of powerful opponents. It breathes fire of such great heat that it melts anything."
+gen_neural pikachu "Pikachu. Mouse pokemon. When several of these pokemon gather, their electricity could build and cause lightning storms."
+
+# ---- Stage 3: Piper neural TTS, cadence-fixed pass ----
+# Fix = (a) standard "X, the Y Pokemon." phrasing instead of two short
+# sentences, which stopped "Pokemon" from landing phrase-final where
+# duration models over-lengthen it, and (b) reduced noise-w-scale (duration
+# randomness) and noise-scale (audio variance) so rare proper nouns don't
+# get a random stretched/warped rendering.
+gen_fixed() {
+  local name="$1" text="$2"
+  echo "$text" | piper -m "$VOICE_MODEL" -f "${name}_fixed.wav" \
+    --noise-scale 0.5 --noise-w-scale 0.3 --length-scale 0.98
+}
+gen_fixed bulbasaur "Bulbasaur, the seed Pokémon. It can be seen napping in bright sunlight. There is a seed on its back. By soaking up the sun's rays, the seed grows progressively larger."
+gen_fixed charizard "Charizard, the flame Pokémon. Charizard flies around the sky in search of powerful opponents. It breathes fire of such great heat that it melts anything."
+gen_fixed pikachu "Pikachu, the mouse Pokémon. When several of these Pokémon gather, their electricity could build and cause lightning storms."
+
+# ---- Stage 4: robot-voice DSP filter passes ----
+# robot_filter.py         -> heavy/original (square-wave ring mod + bitcrush + narrow bandpass)
+# robot_filter_light.py   -> barely-there (sine ring mod, wide bandpass, mild drive)
+# robot_filter_v2.py      -> parameterized 0..1 intensity knob (used at 0.45 = "medium")
+for name in bulbasaur charizard pikachu; do
+  python3 robot_filter.py       "${name}_raw.wav"   "${name}_robot.wav"
+  python3 robot_filter_light.py "${name}_neural.wav" "${name}_neural_light.wav"
+  python3 robot_filter_v2.py    "${name}_fixed.wav"  "${name}_fixed_robot.wav" 0.45
+done
+
+# ---- Stage 5: mp3 for easy playback/delivery ----
+for f in *_raw.wav *_robot.wav *_neural.wav *_neural_light.wav *_fixed.wav *_fixed_robot.wav; do
+  [ -f "$f" ] && ffmpeg -y -loglevel error -i "$f" -codec:a libmp3lame -qscale:a 4 "${f%.wav}.mp3"
+done
+
+echo "Done. See *.mp3 for output."

saved/pokedex_voice_spike_scripts/robot_filter.py

@@ -0,0 +1,79 @@
+"""
+Quick-and-dirty 'robot voice' post-processor for the Pokedex voice spike.
+
+Pipeline: TTS wav in -> ring modulation + bitcrush + band-pass (speaker-grille
+coloring) + light hard-clip drive -> wav out.
+
+Usage: python3 robot_filter.py in.wav out.wav
+"""
+import sys
+import numpy as np
+from scipy.io import wavfile
+from scipy.signal import butter, sosfilt
+
+
+def load(path):
+    sr, data = wavfile.read(path)
+    if data.dtype == np.int16:
+        x = data.astype(np.float64) / 32768.0
+    elif data.dtype == np.int32:
+        x = data.astype(np.float64) / 2147483648.0
+    elif data.dtype == np.uint8:
+        x = (data.astype(np.float64) - 128) / 128.0
+    else:
+        x = data.astype(np.float64)
+    if x.ndim > 1:
+        x = x.mean(axis=1)
+    return sr, x
+
+
+def save(path, sr, x):
+    x = np.clip(x, -1.0, 1.0)
+    wavfile.write(path, sr, (x * 32767).astype(np.int16))
+
+
+def ring_modulate(x, sr, carrier_hz=45.0, mix=0.55):
+    t = np.arange(len(x)) / sr
+    carrier = np.sign(np.sin(2 * np.pi * carrier_hz * t))  # square carrier = harsher/metallic
+    modulated = x * carrier
+    return (1 - mix) * x + mix * modulated
+
+
+def bitcrush(x, bit_depth=6, sr=None, downsample_factor=3):
+    # sample-and-hold downsample (classic lo-fi/robotic stepping)
+    if downsample_factor > 1:
+        held = np.repeat(x[::downsample_factor], downsample_factor)[: len(x)]
+        if len(held) < len(x):
+            held = np.pad(held, (0, len(x) - len(held)))
+        x = held
+    levels = 2 ** bit_depth
+    x = np.round(x * levels) / levels
+    return x
+
+
+def bandpass(x, sr, low=350.0, high=3400.0, order=4):
+    sos = butter(order, [low, high], btype="band", fs=sr, output="sos")
+    return sosfilt(sos, x)
+
+
+def drive(x, amount=1.8):
+    return np.tanh(x * amount) / np.tanh(amount)
+
+
+def robotize(x, sr):
+    y = ring_modulate(x, sr, carrier_hz=45.0, mix=0.5)
+    y = bitcrush(y, bit_depth=7, downsample_factor=2)
+    y = bandpass(y, sr, low=300, high=3800)
+    y = drive(y, amount=1.6)
+    # normalize
+    peak = np.max(np.abs(y)) or 1.0
+    y = y / peak * 0.9
+    return y
+
+
+if __name__ == "__main__":
+    src, dst = sys.argv[1], sys.argv[2]
+    sr, x = load(src)
+    y = robotize(x, sr)
+    save(dst, sr, y)
+    print(f"wrote {dst}")

saved/pokedex_voice_spike_scripts/robot_filter_light.py

@@ -0,0 +1,58 @@
+"""
+Light-touch robot coloring for a neural (already-natural-sounding) TTS voice.
+Goal: keep natural cadence/prosody intact, just add a synthesized/device edge.
+"""
+import sys
+import numpy as np
+from scipy.io import wavfile
+from scipy.signal import butter, sosfilt
+
+
+def load(path):
+    sr, data = wavfile.read(path)
+    if data.dtype == np.int16:
+        x = data.astype(np.float64) / 32768.0
+    elif data.dtype == np.int32:
+        x = data.astype(np.float64) / 2147483648.0
+    else:
+        x = data.astype(np.float64)
+    if x.ndim > 1:
+        x = x.mean(axis=1)
+    return sr, x
+
+
+def save(path, sr, x):
+    x = np.clip(x, -1.0, 1.0)
+    wavfile.write(path, sr, (x * 32767).astype(np.int16))
+
+
+def ring_modulate(x, sr, carrier_hz=90.0, mix=0.12):
+    t = np.arange(len(x)) / sr
+    carrier = np.sin(2 * np.pi * carrier_hz * t)  # sine carrier = subtler than square
+    return (1 - mix) * x + mix * (x * carrier)
+
+
+def bandpass(x, sr, low=200.0, high=5500.0, order=2):
+    sos = butter(order, [low, high], btype="band", fs=sr, output="sos")
+    return sosfilt(sos, x)
+
+
+def drive(x, amount=1.15):
+    return np.tanh(x * amount) / np.tanh(amount)
+
+
+def robotize_light(x, sr):
+    y = ring_modulate(x, sr, carrier_hz=90.0, mix=0.12)
+    y = bandpass(y, sr, low=200, high=5500)
+    y = drive(y, amount=1.15)
+    peak = np.max(np.abs(y)) or 1.0
+    y = y / peak * 0.9
+    return y
+
+
+if __name__ == "__main__":
+    src, dst = sys.argv[1], sys.argv[2]
+    sr, x = load(src)
+    y = robotize_light(x, sr)
+    save(dst, sr, y)
+    print(f"wrote {dst}")

saved/pokedex_voice_spike_scripts/robot_filter_v2.py

@@ -0,0 +1,72 @@
+"""
+Parameterized robot-voice filter, v2.
+intensity in [0, 1]: 0 = untouched, 1 = full "heavy" robot from round 1.
+"""
+import sys
+import numpy as np
+from scipy.io import wavfile
+from scipy.signal import butter, sosfilt
+
+
+def load(path):
+    sr, data = wavfile.read(path)
+    if data.dtype == np.int16:
+        x = data.astype(np.float64) / 32768.0
+    elif data.dtype == np.int32:
+        x = data.astype(np.float64) / 2147483648.0
+    else:
+        x = data.astype(np.float64)
+    if x.ndim > 1:
+        x = x.mean(axis=1)
+    return sr, x
+
+
+def save(path, sr, x):
+    x = np.clip(x, -1.0, 1.0)
+    wavfile.write(path, sr, (x * 32767).astype(np.int16))
+
+
+def ring_modulate(x, sr, carrier_hz, mix, square=False):
+    t = np.arange(len(x)) / sr
+    carrier = np.sign(np.sin(2 * np.pi * carrier_hz * t)) if square else np.sin(2 * np.pi * carrier_hz * t)
+    return (1 - mix) * x + mix * (x * carrier)
+
+
+def bitcrush(x, bit_depth, downsample_factor):
+    if downsample_factor > 1:
+        held = np.repeat(x[::downsample_factor], downsample_factor)[: len(x)]
+        if len(held) < len(x):
+            held = np.pad(held, (0, len(x) - len(held)))
+        x = held
+    levels = 2 ** bit_depth
+    return np.round(x * levels) / levels
+
+
+def bandpass(x, sr, low, high, order=3):
+    sos = butter(order, [low, high], btype="band", fs=sr, output="sos")
+    return sosfilt(sos, x)
+
+
+def drive(x, amount):
+    return np.tanh(x * amount) / np.tanh(amount)
+
+
+def robotize(x, sr, intensity=0.45):
+    i = intensity
+    y = ring_modulate(x, sr, carrier_hz=60 - 15 * i, mix=0.1 + 0.45 * i, square=(i > 0.6))
+    if i > 0.35:
+        y = bitcrush(y, bit_depth=int(12 - 6 * i), downsample_factor=1 if i < 0.7 else 2)
+    y = bandpass(y, sr, low=250 - 100 * i, high=6000 - 2200 * i)
+    y = drive(y, amount=1.0 + 0.9 * i)
+    peak = np.max(np.abs(y)) or 1.0
+    y = y / peak * 0.9
+    return y
+
+
+if __name__ == "__main__":
+    src, dst = sys.argv[1], sys.argv[2]
+    intensity = float(sys.argv[3]) if len(sys.argv) > 3 else 0.45
+    sr, x = load(src)
+    y = robotize(x, sr, intensity=intensity)
+    save(dst, sr, y)
+    print(f"wrote {dst} (intensity={intensity})")

saved/pokedex_voice_spike_scripts/voices/en_US-joe-medium.onnx.json

@@ -0,0 +1,484 @@
+{
+  "audio": {
+    "sample_rate": 22050,
+    "quality": "medium"
+  },
+  "espeak": {
+    "voice": "en-us"
+  },
+  "inference": {
+    "noise_scale": 0.667,
+    "length_scale": 1,
+    "noise_w": 0.8
+  },
+  "phoneme_type": "espeak",
+  "phoneme_map": {},
+  "phoneme_id_map": {
+    "_": [
+      0
+    ],
+    "^": [
+      1
+    ],
+    "$": [
+      2
+    ],
+    " ": [
+      3
+    ],
+    "!": [
+      4
+    ],
+    "'": [
+      5
+    ],
+    "(": [
+      6
+    ],
+    ")": [
+      7
+    ],
+    ",": [
+      8
+    ],
+    "-": [
+      9
+    ],
+    ".": [
+      10
+    ],
+    ":": [
+      11
+    ],
+    ";": [
+      12
+    ],
+    "?": [
+      13
+    ],
+    "a": [
+      14
+    ],
+    "b": [
+      15
+    ],
+    "c": [
+      16
+    ],
+    "d": [
+      17
+    ],
+    "e": [
+      18
+    ],
+    "f": [
+      19
+    ],
+    "h": [
+      20
+    ],
+    "i": [
+      21
+    ],
+    "j": [
+      22
+    ],
+    "k": [
+      23
+    ],
+    "l": [
+      24
+    ],
+    "m": [
+      25
+    ],
+    "n": [
+      26
+    ],
+    "o": [
+      27
+    ],
+    "p": [
+      28
+    ],
+    "q": [
+      29
+    ],
+    "r": [
+      30
+    ],
+    "s": [
+      31
+    ],
+    "t": [
+      32
+    ],
+    "u": [
+      33
+    ],
+    "v": [
+      34
+    ],
+    "w": [
+      35
+    ],
+    "x": [
+      36
+    ],
+    "y": [
+      37
+    ],
+    "z": [
+      38
+    ],
+    "æ": [
+      39
+    ],
+    "ç": [
+      40
+    ],
+    "ð": [
+      41
+    ],
+    "ø": [
+      42
+    ],
+    "ħ": [
+      43
+    ],
+    "ŋ": [
+      44
+    ],
+    "œ": [
+      45
+    ],
+    "ǀ": [
+      46
+    ],
+    "ǁ": [
+      47
+    ],
+    "ǂ": [
+      48
+    ],
+    "ǃ": [
+      49
+    ],
+    "ɐ": [
+      50
+    ],
+    "ɑ": [
+      51
+    ],
+    "ɒ": [
+      52
+    ],
+    "ɓ": [
+      53
+    ],
+    "ɔ": [
+      54
+    ],
+    "ɕ": [
+      55
+    ],
+    "ɖ": [
+      56
+    ],
+    "ɗ": [
+      57
+    ],
+    "ɘ": [
+      58
+    ],
+    "ə": [
+      59
+    ],
+    "ɚ": [
+      60
+    ],
+    "ɛ": [
+      61
+    ],
+    "ɜ": [
+      62
+    ],
+    "ɞ": [
+      63
+    ],
+    "ɟ": [
+      64
+    ],
+    "ɠ": [
+      65
+    ],
+    "ɡ": [
+      66
+    ],
+    "ɢ": [
+      67
+    ],
+    "ɣ": [
+      68
+    ],
+    "ɤ": [
+      69
+    ],
+    "ɥ": [
+      70
+    ],
+    "ɦ": [
+      71
+    ],
+    "ɧ": [
+      72
+    ],
+    "ɨ": [
+      73
+    ],
+    "ɪ": [
+      74
+    ],
+    "ɫ": [
+      75
+    ],
+    "ɬ": [
+      76
+    ],
+    "ɭ": [
+      77
+    ],
+    "ɮ": [
+      78
+    ],
+    "ɯ": [
+      79
+    ],
+    "ɰ": [
+      80
+    ],
+    "ɱ": [
+      81
+    ],
+    "ɲ": [
+      82
+    ],
+    "ɳ": [
+      83
+    ],
+    "ɴ": [
+      84
+    ],
+    "ɵ": [
+      85
+    ],
+    "ɶ": [
+      86
+    ],
+    "ɸ": [
+      87
+    ],
+    "ɹ": [
+      88
+    ],
+    "ɺ": [
+      89
+    ],
+    "ɻ": [
+      90
+    ],
+    "ɽ": [
+      91
+    ],
+    "ɾ": [
+      92
+    ],
+    "ʀ": [
+      93
+    ],
+    "ʁ": [
+      94
+    ],
+    "ʂ": [
+      95
+    ],
+    "ʃ": [
+      96
+    ],
+    "ʄ": [
+      97
+    ],
+    "ʈ": [
+      98
+    ],
+    "ʉ": [
+      99
+    ],
+    "ʊ": [
+      100
+    ],
+    "ʋ": [
+      101
+    ],
+    "ʌ": [
+      102
+    ],
+    "ʍ": [
+      103
+    ],
+    "ʎ": [
+      104
+    ],
+    "ʏ": [
+      105
+    ],
+    "ʐ": [
+      106
+    ],
+    "ʑ": [
+      107
+    ],
+    "ʒ": [
+      108
+    ],
+    "ʔ": [
+      109
+    ],
+    "ʕ": [
+      110
+    ],
+    "ʘ": [
+      111
+    ],
+    "ʙ": [
+      112
+    ],
+    "ʛ": [
+      113
+    ],
+    "ʜ": [
+      114
+    ],
+    "ʝ": [
+      115
+    ],
+    "ʟ": [
+      116
+    ],
+    "ʡ": [
+      117
+    ],
+    "ʢ": [
+      118
+    ],
+    "ʲ": [
+      119
+    ],
+    "ˈ": [
+      120
+    ],
+    "ˌ": [
+      121
+    ],
+    "ː": [
+      122
+    ],
+    "ˑ": [
+      123
+    ],
+    "˞": [
+      124
+    ],
+    "β": [
+      125
+    ],
+    "θ": [
+      126
+    ],
+    "χ": [
+      127
+    ],
+    "ᵻ": [
+      128
+    ],
+    "ⱱ": [
+      129
+    ],
+    "0": [
+      130
+    ],
+    "1": [
+      131
+    ],
+    "2": [
+      132
+    ],
+    "3": [
+      133
+    ],
+    "4": [
+      134
+    ],
+    "5": [
+      135
+    ],
+    "6": [
+      136
+    ],
+    "7": [
+      137
+    ],
+    "8": [
+      138
+    ],
+    "9": [
+      139
+    ],
+    "̧": [
+      140
+    ],
+    "̃": [
+      141
+    ],
+    "̪": [
+      142
+    ],
+    "̯": [
+      143
+    ],
+    "̩": [
+      144
+    ],
+    "ʰ": [
+      145
+    ],
+    "ˤ": [
+      146
+    ],
+    "ε": [
+      147
+    ],
+    "↓": [
+      148
+    ],
+    "#": [
+      149
+    ],
+    "\"": [
+      150
+    ]
+  },
+  "num_symbols": 256,
+  "num_speakers": 1,
+  "speaker_id_map": {},
+  "piper_version": "1.0.0",
+  "language": {
+    "code": "en_US",
+    "family": "en",
+    "region": "US",
+    "name_native": "English",
+    "name_english": "English",
+    "country_english": "United States"
+  },
+  "dataset": "joe"
+}