FIRST-PARTY EVIDENCE · 14 SEP 2026

Hear the result. Inspect the evidence.

Original inputs, stored audio and costs across two dated runs. Completed, pending and unstarted cases are reported separately; a planned matrix is not completed coverage.

Listen to outputs ↓Download machine-readable report ↓

Earlier baseline · clone-v1 · Rowan · MP3

23successful stored generations

$0.004984actual generation charge

623billed characters

4sampled languages

Internal example, not a customer case or independent benchmark. These costs are the historical API charges, not our GPU costs.

A three-voice, ten-language plan. Actual results below.

29/30generated cases

1terminal failures retained

$0.013408actual API charge

29/29automatic language matches

All 30 planned cases have a terminal result: 29 generated, 1 failed. The interrupted Lily/French request was fenced before generation admission and its reservation released, with no duplicate generation. It remains a failure in this report, not an all-pass claim.

Planned: three published voices with English reference recordings, one short learning-app prompt per language, serial submission. All submitted cases used the China region; no US performance claim is made. Only saved outputs have a player. Playback creates no new generation charges.

Important: decoded audio and ASR language matches are technical checks, not native-speaker or naturalness ratings. Transcripts can be wrong. Human pronunciation and speaker-similarity review is still pending. Timings include queue polling, transport and download.

qwen3-tts-0.6b-base-nari-shared-v4 · 2026-09-14

Measured time to saved audio

29successful timed cases

71.5smedian wall time

129.8sp95 wall time

China region, serial short prompts; nearest-rank percentiles. Includes queueing, polling, network and download. This is not first-audio latency, model inference time, a load test or an SLA. The failed case is retained separately and excluded from success timings.

Review pronunciation, naturalness and similarity →

Download all inputs, hashes and results ↓

English · 3 / 3 outputs saved

Rowan

Welcome back. Your next lesson starts at nine. Are you ready to begin?

5.98s audio · 69.1s wall time · $0.000560

Download MP3 ↓

Automatic transcript · en ✓

Welcome back. Your next lesson starts at 9. Are you ready to begin?

Machine output, not a human verdict.

Lily

Welcome back. Your next lesson starts at nine. Are you ready to begin?

4.20s audio · 134.0s wall time · $0.000560

Download MP3 ↓

Automatic transcript · en ✓

Welcome back. Your next lesson starts at 9. Are you ready to begin?

Machine output, not a human verdict.

Owen

Welcome back. Your next lesson starts at nine. Are you ready to begin?

3.65s audio · 89.2s wall time · $0.000560

Download MP3 ↓

Automatic transcript · en ✓

Welcome back, your next lesson starts at 9. Are you ready to begin?

Machine output, not a human verdict.

中文 · 3 / 3 outputs saved

Rowan

欢迎回来。下一节课九点开始，你准备好了吗？我们一起学习新的知识。

6.77s audio · 63.5s wall time · $0.000256

Download MP3 ↓

Automatic transcript · zh ✓

欢迎回来,下一节课9点开始,你准备好了吗?我们一起学习新的知识。

Machine output, not a human verdict.

Lily

欢迎回来。下一节课九点开始，你准备好了吗？我们一起学习新的知识。

7.66s audio · 80.3s wall time · $0.000256

Download MP3 ↓

Automatic transcript · zh ✓

歡迎回來下一節刻9點開始你準備好了嗎?我們一起學習新的知識

Machine output, not a human verdict.

Owen

欢迎回来。下一节课九点开始，你准备好了吗？我们一起学习新的知识。

5.66s audio · 82.8s wall time · $0.000256

Download MP3 ↓

Automatic transcript · zh ✓

欢迎回来下一个课9点开始你准备好了吗我们一起学习新的知识

Machine output, not a human verdict.

Deutsch · 3 / 3 outputs saved

Rowan

Willkommen zurück. Die nächste Stunde beginnt um neun. Bist du bereit?

4.85s audio · 59.0s wall time · $0.000560

Download MP3 ↓

Automatic transcript · de ✓

Willkommen zurück. Die nächste Stunde beginnt um 9. Bist du bereit?

Machine output, not a human verdict.

Lily

Willkommen zurück. Die nächste Stunde beginnt um neun. Bist du bereit?

4.20s audio · 82.7s wall time · $0.000560

Download MP3 ↓

Automatic transcript · de ✓

Willkommen zurück! Die nächste Stunde beginnt um 9. Bist du bereit?

Machine output, not a human verdict.

Owen

Willkommen zurück. Die nächste Stunde beginnt um neun. Bist du bereit?

4.54s audio · 61.1s wall time · $0.000560

Download MP3 ↓

Automatic transcript · de ✓

Willkommen zurück! Die nächste Stunde beginnt um 9. Wirst du bereit?

Machine output, not a human verdict.

日本語 · 3 / 3 outputs saved

Rowan

お帰りなさい。次の授業は九時に始まります。準備はできましたか。

5.50s audio · 72.1s wall time · $0.000248

Download MP3 ↓

Automatic transcript · ja ✓

お帰りなさい次の授業は9時に始まります準備はできましたか

Machine output, not a human verdict.

Lily

お帰りなさい。次の授業は九時に始まります。準備はできましたか。

7.01s audio · 82.4s wall time · $0.000248

Download MP3 ↓

Automatic transcript · ja ✓

お帰りなさい。次の授業は、9時に始まります。準備はできましたか?

Machine output, not a human verdict.

Owen

お帰りなさい。次の授業は九時に始まります。準備はできましたか。

5.47s audio · 61.7s wall time · $0.000248

Download MP3 ↓

Automatic transcript · ja ✓

お帰りなさい次の授業は9時に始まります準備はできましたか

Machine output, not a human verdict.

Français · 2 / 3 outputs saved

Rowan

Bon retour. Votre prochain cours commence à neuf heures. Êtes-vous prêt ?

4.70s audio · 65.5s wall time · $0.000584

Download MP3 ↓

Automatic transcript · fr ✓

Bon retour ! Votre prochain cours commence à 9h, êtes-vous prêt ?

Machine output, not a human verdict.

Lily

Bon retour. Votre prochain cours commence à neuf heures. Êtes-vous prêt ?

Generation failed; recovery completed. Reserved balance released, $0 charged, no replacement generated for this case.

Owen

Bon retour. Votre prochain cours commence à neuf heures. Êtes-vous prêt ?

5.33s audio · 67.4s wall time · $0.000584

Download MP3 ↓

Automatic transcript · fr ✓

Bon retour ! Votre prochain cours commence à 9h. Et de vous prêt ?

Machine output, not a human verdict.

Español · 3 / 3 outputs saved

Rowan

Bienvenido de nuevo. Tu próxima clase empieza a las nueve. ¿Estás listo?

5.66s audio · 66.0s wall time · $0.000576

Download MP3 ↓

Automatic transcript · es ✓

Bienvenido de nuevo. Tu próxima clase empieza a las 9. ¿Estás listo?

Machine output, not a human verdict.

Lily

Bienvenido de nuevo. Tu próxima clase empieza a las nueve. ¿Estás listo?

5.40s audio · 86.7s wall time · $0.000576

Download MP3 ↓

Automatic transcript · es ✓

Bienvenido de nuevo. Tu próxima clase empieza a las 9. ¿Estás listo?

Machine output, not a human verdict.

Owen

Bienvenido de nuevo. Tu próxima clase empieza a las nueve. ¿Estás listo?

5.50s audio · 75.7s wall time · $0.000576

Download MP3 ↓

Automatic transcript · es ✓

Bienvenido de nuevo. Tu próxima clase empieza a las 9. ¿Estás listo?

Machine output, not a human verdict.

한국어 · 3 / 3 outputs saved

Rowan

다시 오신 것을 환영합니다. 다음 수업은 아홉 시에 시작합니다. 준비되셨나요?

5.86s audio · 80.6s wall time · $0.000344

Download MP3 ↓

Automatic transcript · ko ✓

다시 오신 것을 환영합니다. 다음 수업은 9시에 시작합니다. 준비 되셨나요.

Machine output, not a human verdict.

Lily

다시 오신 것을 환영합니다. 다음 수업은 아홉 시에 시작합니다. 준비되셨나요?

6.84s audio · 67.8s wall time · $0.000344

Download MP3 ↓

Automatic transcript · ko ✓

다시 오신 것을 환영합니다. 다음 수업은 9시에 시작합니다. 준비되셨나요?

Machine output, not a human verdict.

Owen

다시 오신 것을 환영합니다. 다음 수업은 아홉 시에 시작합니다. 준비되셨나요?

6.38s audio · 68.2s wall time · $0.000344

Download MP3 ↓

Automatic transcript · ko ✓

다시 오신 것을 환영합니다. 다음 수업은 9시에 시작합니다. 준비되셨나요?

Machine output, not a human verdict.

Português · 3 / 3 outputs saved

Rowan

Bem-vindo de volta. A próxima aula começa às nove horas. Está pronto?

4.68s audio · 61.7s wall time · $0.000552

Download MP3 ↓

Automatic transcript · pt ✓

Bem-vindo de volta. A próxima aula começa as nove horas. Está pronto!

Machine output, not a human verdict.

Lily

Bem-vindo de volta. A próxima aula começa às nove horas. Está pronto?

5.64s audio · 63.5s wall time · $0.000552

Download MP3 ↓

Automatic transcript · pt ✓

Bem-vindo de volta. A próxima aula começa às 9 horas. Está pronto?

Machine output, not a human verdict.

Owen

Bem-vindo de volta. A próxima aula começa às nove horas. Está pronto?

4.92s audio · 129.8s wall time · $0.000552

Download MP3 ↓

Automatic transcript · pt ✓

Bem-vindo de volta. A próxima aula começa as nove horas. Está pronto?

Machine output, not a human verdict.

Русский · 3 / 3 outputs saved

Rowan

С возвращением. Следующий урок начинается в девять. Вы готовы?

5.64s audio · 59.9s wall time · $0.000496

Download MP3 ↓

Automatic transcript · ru ✓

С возвращением. Следующий урок начинается в 9. Вы готовы?

Machine output, not a human verdict.

Lily

С возвращением. Следующий урок начинается в девять. Вы готовы?

5.16s audio · 75.4s wall time · $0.000496

Download MP3 ↓

Automatic transcript · ru ✓

Звращением. Следующий урок начинается в 9. Вы готовы?

Machine output, not a human verdict.

Owen

С возвращением. Следующий урок начинается в девять. Вы готовы?

5.50s audio · 80.8s wall time · $0.000496

Download MP3 ↓

Automatic transcript · ru ✓

С возвращением. Следующий урок начинается в 9. Вы готовы?

Machine output, not a human verdict.

Italiano · 3 / 3 outputs saved

Rowan

Bentornato. La prossima lezione inizia alle nove. Sei pronto?

4.94s audio · 71.5s wall time · $0.000488

Download MP3 ↓

Automatic transcript · it ✓

Ben tornato. La prossima lezione inizia alle nove. Sei pronto?

Machine output, not a human verdict.

Lily

Bentornato. La prossima lezione inizia alle nove. Sei pronto?

4.61s audio · 74.3s wall time · $0.000488

Download MP3 ↓

Automatic transcript · it ✓

Ben tornato! La prossima lezione inizia le 9. Sei pronto?

Machine output, not a human verdict.

Owen

Bentornato. La prossima lezione inizia alle nove. Sei pronto?

5.16s audio · 65.1s wall time · $0.000488

Download MP3 ↓

Automatic transcript · it ✓

Ben tornato! La prossima lezione inizia alle nove! Sei pronto?

Machine output, not a human verdict.

Original text stays beside the audio.

EN · Rowan

Every small step opens a new door. Listen, learn, and tell your own story.

5.136s · 41,709 bytes · $0.000592

Download MP3 ↓

ZH · Rowan

每一个小小的进步，都会打开一扇新的门。认真聆听，不断学习，讲述属于你自己的故事。

7.896s · 63,213 bytes · $0.000320

Download MP3 ↓

JA · Rowan

小さな一歩が、新しい扉を開きます。耳を傾け、学び、自分だけの物語を語りましょう。

7.416s · 59,373 bytes · $0.000320

Download MP3 ↓

Texts, hashes and usage manifest ↗Compare the original reference →

Method and reproducibility

Use one operator-published Rowan preset: three comparison texts and 20 cards across en, zh, ja and es. Execute ordinary paid API jobs; no trial characters used.
Save outputs with the original input, returned usage, size, duration and SHA-256. Decode all 23 MP3 files and check the hash-to-file association.
Inspect 14 measured English word intervals. The player uses actual audio time, preserves gaps, and does not invent word boundaries.
Run the complete saved 20-card project again. Matching local checksums skip API calls. Download and run the starter to inspect recovery behavior yourself.

Runnable project and full procedure →Measured timestamp player →All 20 card files + manifest ↓

What this report does not prove

It does not measure service-wide success rate: this manifest records completed outputs, not every attempt or live traffic.
One voice and four sampled languages do not verify every voice or all configured languages.
Offline ASR detected the expected language, but some Chinese and Japanese transcripts differed from the input. No native-speaker review or universal naturalness/similarity score is claimed.
No latency percentiles for this earlier baseline, load test, independent competitor comparison or SLA. A new generation may sound different and will not match these file hashes.

A practical acceptance checklist for your app

Use your own representative scripts and authorized voice. Compare pronunciation, names, numbers and similarity with your target users. Record request-to-download time separately from voice preparation and queue waiting. Save actual billed characters and repeat-generation counts when assessing cost.

Apply the integration checklist →Estimate your text →