Datasets:

DograhAI-hq
/

tts-verbalization-showcase

item_id stringclasses 4 values	model stringclasses 7 values	model_revision stringclasses 7 values	category stringclasses 4 values	written stringclasses 4 values	spoken_reference stringclasses 4 values	target_span stringclasses 4 values	whisper_transcript stringlengths 40 64	whisper_correct bool 2 classes	nova3_transcript stringlengths 39 110	nova3_correct bool 2 classes
syn-en-currency-001	chatterbox	chatterbox@1b8422bc	Currency	They raised $4,825 for the local hospital fund.	They raised four thousand eight hundred twenty five dollars for the local hospital fund.	four thousand eight hundred twenty five dollars	They raised $4,825 for the local hospital fund.	true	they raised four thousand eight hundred twenty five for the local hospital fund	false
syn-en-currency-001	chatterbox-multilingual	chatterbox-multilingual@9cfba4c2	Currency	They raised $4,825 for the local hospital fund.	They raised four thousand eight hundred twenty five dollars for the local hospital fund.	four thousand eight hundred twenty five dollars	They raised $4,695 for the local hospital fund.	false	they raised four hundred sixty thousand one hundred and ninety five for the local hospital fund	false
syn-en-currency-001	chatterbox-turbo	chatterbox-turbo@95c87b88	Currency	They raised $4,825 for the local hospital fund.	They raised four thousand eight hundred twenty five dollars for the local hospital fund.	four thousand eight hundred twenty five dollars	They raised $482,825 for the local hospital fund.	false	they raised four hundred eighty two thousand eight hundred and twenty five dollars for the local hospital fund	false
syn-en-currency-001	kokoro-82m	kokoro-82m@f559560e	Currency	They raised $4,825 for the local hospital fund.	They raised four thousand eight hundred twenty five dollars for the local hospital fund.	four thousand eight hundred twenty five dollars	They raised $4,825 for the local hospital fund.	true	they raised four thousand eight hundred and twenty five dollars for the local hospital fund	true
syn-en-currency-001	orpheus-3b	orpheus-3b-0.1-ft@79f2a473	Currency	They raised $4,825 for the local hospital fund.	They raised four thousand eight hundred twenty five dollars for the local hospital fund.	four thousand eight hundred twenty five dollars	They raised $4,825 for the local hospital fund.	true	they raised four thousand eight hundred and twenty five dollars for the local hospital fund	true
syn-en-currency-001	parler-tts	parler-tts@bf38249a+desc-clear	Currency	They raised $4,825 for the local hospital fund.	They raised four thousand eight hundred twenty five dollars for the local hospital fund.	four thousand eight hundred twenty five dollars	They raised WACA for the local hospital fund.	false	they raised warcub for the local hospital fund	false
syn-en-currency-001	qwen3-tts	qwen3-tts@d490a561	Currency	They raised $4,825 for the local hospital fund.	They raised four thousand eight hundred twenty five dollars for the local hospital fund.	four thousand eight hundred twenty five dollars	They raised $4,825 for the local hospital fund.	true	they raised four thousand eight hundred and twenty five dollars for the local hospital fund	true
syn-en-date-001	chatterbox	chatterbox@1b8422bc	Date	Her birthday falls on 2009-02-26 this year.	Her birthday falls on February twenty sixth two thousand nine this year.	february twenty sixth two thousand nine	Her birthday falls on 2009-02-06 this year.	false	her birthday falls on two thousand nine two two six this year	false
syn-en-date-001	chatterbox-multilingual	chatterbox-multilingual@9cfba4c2	Date	Her birthday falls on 2009-02-26 this year.	Her birthday falls on February twenty sixth two thousand nine this year.	february twenty sixth two thousand nine	Her birthday falls on 2009 NIV 226 this year.	false	her birthday falls on two thousand and nine nif-two twenty six this year	false
syn-en-date-001	chatterbox-turbo	chatterbox-turbo@95c87b88	Date	Her birthday falls on 2009-02-26 this year.	Her birthday falls on February twenty sixth two thousand nine this year.	february twenty sixth two thousand nine	Her birthday falls on 2009-02-26 this year.	false	her birthday falls on two thousand and nine two twenty six this year	false
syn-en-date-001	kokoro-82m	kokoro-82m@f559560e	Date	Her birthday falls on 2009-02-26 this year.	Her birthday falls on February twenty sixth two thousand nine this year.	february twenty sixth two thousand nine	Her birthday falls on 2090, 2-26 this year.	false	her birthday falls on two thousand nine zero two twenty six this year	false
syn-en-date-001	orpheus-3b	orpheus-3b-0.1-ft@79f2a473	Date	Her birthday falls on 2009-02-26 this year.	Her birthday falls on February twenty sixth two thousand nine this year.	february twenty sixth two thousand nine	Her birthday falls on 2009-02-26 this year.	false	her birthday falls on two thousand and nine oh two twenty six this year	false
syn-en-date-001	parler-tts	parler-tts@bf38249a+desc-clear	Date	Her birthday falls on 2009-02-26 this year.	Her birthday falls on February twenty sixth two thousand nine this year.	february twenty sixth two thousand nine	Herbal tea flowers enswore so, very gee. Thok!	false	her birthday fais and voslo vargy vogue	false
syn-en-date-001	qwen3-tts	qwen3-tts@d490a561	Date	Her birthday falls on 2009-02-26 this year.	Her birthday falls on February twenty sixth two thousand nine this year.	february twenty sixth two thousand nine	Her birthday falls on 2009.06 this year.	false	her birthday falls on two thousand and nine dot r six this year	false
syn-en-hashtag-or-mention-001	chatterbox	chatterbox@1b8422bc	Hashtag or Mention	Follow #GreenBreak for updates during the conference.	Follow hashtag green break for updates during the conference.	hashtag green break	Follow Green Break for updates during the conference.	false	follow greenbreak for updates during the conference	false
syn-en-hashtag-or-mention-001	chatterbox-multilingual	chatterbox-multilingual@9cfba4c2	Hashtag or Mention	Follow #GreenBreak for updates during the conference.	Follow hashtag green break for updates during the conference.	hashtag green break	Follow for Green Break for updates during the conference.	false	follow ferc green break for updates during the conference	false
syn-en-hashtag-or-mention-001	chatterbox-turbo	chatterbox-turbo@95c87b88	Hashtag or Mention	Follow #GreenBreak for updates during the conference.	Follow hashtag green break for updates during the conference.	hashtag green break	Follow numbergreenbreak for updates during the conference.	false	follow number green break for updates during the conference	false
syn-en-hashtag-or-mention-001	kokoro-82m	kokoro-82m@f559560e	Hashtag or Mention	Follow #GreenBreak for updates during the conference.	Follow hashtag green break for updates during the conference.	hashtag green break	Follow GreenBreak for updates during the conference.	false	follow green break for updates during the conference	false
syn-en-hashtag-or-mention-001	orpheus-3b	orpheus-3b-0.1-ft@79f2a473	Hashtag or Mention	Follow #GreenBreak for updates during the conference.	Follow hashtag green break for updates during the conference.	hashtag green break	Follow GreenBreak for updates during the conference.	false	follow green break for updates during the conference	false
syn-en-hashtag-or-mention-001	parler-tts	parler-tts@bf38249a+desc-clear	Hashtag or Mention	Follow #GreenBreak for updates during the conference.	Follow hashtag green break for updates during the conference.	hashtag green break	Palo Verdean Green, Rake for Iron Mandar, during the conference.	false	polowadijk green break for ironman during the conference	false
syn-en-hashtag-or-mention-001	qwen3-tts	qwen3-tts@d490a561	Hashtag or Mention	Follow #GreenBreak for updates during the conference.	Follow hashtag green break for updates during the conference.	hashtag green break	Follow Grand Break for updates during the conference.	false	follow groundbreak for updates during the conference	false
syn-en-musical-notation-001	chatterbox	chatterbox@1b8422bc	Musical Notation	The guitarist opens the piece with G6 every night.	The guitarist opens the piece with G six every night.	g six	The guitarist opens the piece with G6 every night.	true	the guitarist opens a piece with g six every night	true
syn-en-musical-notation-001	chatterbox-multilingual	chatterbox-multilingual@9cfba4c2	Musical Notation	The guitarist opens the piece with G6 every night.	The guitarist opens the piece with G six every night.	g six	The guitarist opens the piece with G.C.s every night.	false	guitarist opens the piece with gcs every night	false
syn-en-musical-notation-001	chatterbox-turbo	chatterbox-turbo@95c87b88	Musical Notation	The guitarist opens the piece with G6 every night.	The guitarist opens the piece with G six every night.	g six	The guitarist opens the piece with G6 every night.	true	the guitarist opens the piece with g6 every night	true
syn-en-musical-notation-001	kokoro-82m	kokoro-82m@f559560e	Musical Notation	The guitarist opens the piece with G6 every night.	The guitarist opens the piece with G six every night.	g six	The guitarist opens the piece with G6 every night.	true	the guitarist opens the piece with g six every night	true
syn-en-musical-notation-001	orpheus-3b	orpheus-3b-0.1-ft@79f2a473	Musical Notation	The guitarist opens the piece with G6 every night.	The guitarist opens the piece with G six every night.	g six	The guitarist opens the keys with G6 every night.	true	the guitarist opens the piece with g6 every night	true
syn-en-musical-notation-001	parler-tts	parler-tts@bf38249a+desc-clear	Musical Notation	The guitarist opens the piece with G6 every night.	The guitarist opens the piece with G six every night.	g six	The guitarist opens the piece with Gawane every night.	false	the guitarist opens the piece with goboli every night	false
syn-en-musical-notation-001	qwen3-tts	qwen3-tts@d490a561	Musical Notation	The guitarist opens the piece with G6 every night.	The guitarist opens the piece with G six every night.	g six	The guitarist opens the piece with G6 every night.	true	the guitarist opens the piece with g six every night	true

TTS Verbalization — showcase clips

Twenty-eight clips: four sentences, each spoken by all seven models. Small enough to embed in a web page, chosen so the four comparisons carry the whole finding.

The complete 7,560-clip set is at DograhAI-hq/tts-verbalization-audio. The corpus, code and results are at dograh-hq/experiments.

The four sentences

1. A date every model gets wrong — syn-en-date-001

Her birthday falls on 2009-02-26 this year.

Correct: "February twenty sixth two thousand nine". All seven fail. This is also the clearest example of parler-tts breaking down: it returned audio transcribed as "Herbal tea flowers enswore so, very gee." Play kokoro and parler back to back.

2. The hashtag zero — syn-en-hashtag-or-mention-001

Follow #GreenBreak for updates during the conference.

Correct: "hashtag green break". Every model reads the words and drops the symbol. Across 279 scored clips in this category, zero are correct, and the word "hashtag" appears in none of the transcripts. The words are split correctly; the symbol is simply never spoken.

3. Currency, where the models separate — syn-en-currency-001

They raised $4,825 for the local hospital fund.

Correct: "four thousand eight hundred twenty five dollars". kokoro-82m gets it, chatterbox-multilingual does not. Currency is the category with the widest spread between models.

4. Something every good model handles — syn-en-musical-notation-001

The guitarist opens the piece with G6 every night.

Correct: "G six". kokoro-82m scores 1.00 on Musical Notation and 0.00 on Hashtag in the same run, which is why a 27-category average is the wrong number to decide on.

Layout

<model>/<item_id>.wav
metadata.jsonl

Each row of metadata.jsonl carries the written text, the reference reading, the scored span, both recognizer transcripts and both verdicts. Field meanings are documented on the full dataset.

Direct playback over HTTP, which is how the write-up embeds these:

<audio controls preload="none"
  src="https://huggingface.co/datasets/DograhAI-hq/tts-verbalization-showcase/resolve/main/kokoro-82m/syn-en-date-001.wav">
</audio>

Licence and provenance

Metadata and corpus text are CC BY 4.0 by Dograh. The audio is output from seven models under Apache 2.0 or MIT, none of which restricts generated audio. Chatterbox clips carry Resemble AI's imperceptible Perth watermark, and orpheus-3b is built on Llama-3.2-3B, which matters if you train on these outputs. Full provenance notes are on the complete dataset card.

Downloads last month: 78

Number of rows:

Total file size:

7.27 MB