audio audioduration (s) 3.88 6.68 | ref_text stringlengths 61 91 | asr_text stringlengths 61 91 | wer float64 0 0 | duration_s float64 3.88 6.68 | pitch_hz int64 83 223 | pitch_band stringclasses 3
values |
|---|---|---|---|---|---|---|
The morning fog rolled across the harbor as the fishing boats prepared to set out. | The morning fog rolled across the harbor as the fishing boats prepared to set out. | 0 | 4.28 | 98 | low | |
She measured each ingredient carefully before folding the batter into the pan. | She measured each ingredient carefully before folding the batter into the pan. | 0 | 5.12 | 117 | low | |
Quantum computers exploit superposition to evaluate many possibilities at once. | Quantum computers exploit superposition to evaluate many possibilities at once. | 0 | 6.68 | 170 | high | |
After the storm finally passed, the children rushed outside to splash in the puddles. | After the storm finally passed, the children rushed outside to splash in the puddles. | 0 | 5.44 | 187 | high | |
The newest exhibit features artifacts recovered from a sunken merchant ship. | The newest exhibit features artifacts recovered from a sunken merchant ship. | 0 | 4.24 | 156 | mid | |
He tightened the last bolt, wiped his hands, and stepped back to admire the engine. | He tightened the last bolt, wiped his hands, and stepped back to admire the engine. | 0 | 5 | 83 | low | |
Migrating geese traced a wide arc across the pale autumn sky. | Migrating geese traced a wide arc across the pale autumn sky. | 0 | 3.96 | 99 | low | |
Our flight was delayed three hours because of unexpected weather over the mountains. | Our flight was delayed three hours because of unexpected weather over the mountains. | 0 | 5.32 | 210 | high | |
The professor explained how coral reefs support nearly a quarter of all marine species. | The professor explained how coral reefs support nearly a quarter of all marine species. | 0 | 6.28 | 175 | high | |
A gentle melody drifted from the open window of the corner cafe. | A gentle melody drifted from the open window of the corner cafe. | 0 | 4.04 | 131 | mid | |
Investors grew nervous as the currency slipped against the dollar for a third straight day. | Investors grew nervous as the currency slipped against the dollar for a third straight day. | 0 | 5 | 180 | high | |
The hikers reached the summit just as the sun broke through the clouds. | The hikers reached the summit just as the sun broke through the clouds. | 0 | 4.4 | 223 | high | |
Please remember to water the tomatoes and lock the gate before you leave. | Please remember to water the tomatoes and lock the gate before you leave. | 0 | 3.92 | 182 | high | |
The ancient manuscript had survived two fires and a flood over six centuries. | The ancient manuscript had survived two fires and a flood over six centuries. | 0 | 5 | 106 | low | |
Engineers tested the new bridge by sending many loaded trucks across it together. | Engineers tested the new bridge by sending many loaded trucks across it together. | 0 | 4.72 | 123 | low | |
My grandmother often told stories about the village where she grew up. | My grandmother often told stories about the village where she grew up. | 0 | 4.32 | 188 | high | |
The robot navigated the crowded warehouse, pausing whenever a worker crossed its path. | The robot navigated the crowded warehouse, pausing whenever a worker crossed its path. | 0 | 6.08 | 174 | high | |
Fresh snow blanketed the rooftops, muffling the usual noise of the city. | Fresh snow blanketed the rooftops, muffling the usual noise of the city. | 0 | 4.64 | 98 | low | |
The committee voted unanimously to fund the new community library. | The committee voted unanimously to fund the new community library. | 0 | 4.16 | 191 | high | |
She practiced the violin every evening until her fingers knew the melody by heart. | She practiced the violin every evening until her fingers knew the melody by heart. | 0 | 5.08 | 173 | high | |
Volcanic ash can disrupt air travel across an entire continent for weeks. | Volcanic ash can disrupt air travel across an entire continent for weeks. | 0 | 4.92 | 115 | low | |
The detective examined the muddy footprints leading away from the garden shed. | The detective examined the muddy footprints leading away from the garden shed. | 0 | 4 | 170 | high | |
We watched dolphins racing alongside the ferry as it crossed the bay. | We watched dolphins racing alongside the ferry as it crossed the bay. | 0 | 3.88 | 143 | mid | |
The startup doubled its revenue after launching the new subscription service. | The startup doubled its revenue after launching the new subscription service. | 0 | 4.12 | 163 | mid | |
A single candle flickered in the window of the old lighthouse cottage. | A single candle flickered in the window of the old lighthouse cottage. | 0 | 4.24 | 184 | high | |
The chef garnished each plate with a sprig of mint and a dusting of cocoa. | The chef garnished each plate with a sprig of mint and a dusting of cocoa. | 0 | 5.04 | 152 | mid | |
Heavy rain is expected throughout the weekend, so bring an umbrella to the game. | Heavy rain is expected throughout the weekend, so bring an umbrella to the game. | 0 | 3.96 | 166 | high | |
Astronomers discovered a faint galaxy that formed shortly after the universe began. | Astronomers discovered a faint galaxy that formed shortly after the universe began. | 0 | 5.28 | 125 | low | |
The puppy chased its tail in circles until it collapsed in a happy heap. | The puppy chased its tail in circles until it collapsed in a happy heap. | 0 | 4.6 | 219 | high | |
Negotiations continued late into the night before the two sides reached an agreement. | Negotiations continued late into the night before the two sides reached an agreement. | 0 | 5.44 | 105 | low |
Higgs-Audio-V3 Voice Bank (30 synthetic reference voices)
30 reference voices for voice-clone TTS benchmarking, generated with bosonai/higgs-audio-v3-tts-4b in default-voice ("smart voice") mode — one clip per random seed (1001-1030), each speaking a distinct enrollment sentence. Use as a controlled, repeatable voice pool instead of per-row dataset reference audios.
Each audio ships with its corresponding reference text (ASR-verified)
Every clip is paired with the transcript of what it actually says. The
reference text was verified by transcribing each clip with
openai/whisper-small and comparing to the generation prompt:
28/30 clips WER = 0.000, mean WER = 0.5% (the two non-zero clips are
number-word normalization, e.g. "forty" vs "40", not real errors). No clip
needed its reference text replaced.
Layout
prompt-wavs/voice_XXXX.wav- 30 reference clips (24 kHz, ~6-8 s each)metadata.csv- audiofolder metadata that drives the HF data viewer:file_name, ref_text, asr_text, wer, duration_s, pitch_hz, pitch_bandmeta.lst-voice_id|ref_text(pipe-delimited; convenience for seed-tts-style loaders)
Voice diversity
Distinct sentence per voice; median pitch spans 83-223 Hz -
9 low (<130 Hz), 5 mid (130-165 Hz), 16 high (>165 Hz). pitch_band in
voices.csv is a coarse voice tag.
Use for voice-clone eval (round-robin)
Pair each target text in your TTS eval set with one of these 30 voices
cyclically. For each request feed prompt-wavs/voice_XXXX.wav as ref_audio
and that voice's ref_text (from meta.lst) as ref_text.
- Downloads last month
- 110