# Video sources and attribution

New clips contain original video with DuraS2ST-generated speech. The source and translated comparison preserve the same encoded video stream. The generated speech is not an original statement by the speaker; no endorsement is implied.

## Chinese → English · Wang Gang introduces Jerry

- Original title: Vlog001 厨师长教外国人做川菜，郭杰瑞表示：这个不辣，还能再辣点吗.webm
- Creator: 美食作家王刚 (Chef Wang Gang)
- Original YouTube: <https://www.youtube.com/watch?v=52Bn8BYFzFg>
- Source and video-specific rights record: <https://commons.wikimedia.org/wiki/File:Vlog001_%E5%8E%A8%E5%B8%88%E9%95%BF%E6%95%99%E5%A4%96%E5%9B%BD%E4%BA%BA%E5%81%9A%E5%B7%9D%E8%8F%9C%EF%BC%8C%E9%83%AD%E6%9D%B0%E7%91%9E%E8%A1%A8%E7%A4%BA%EF%BC%9A%E8%BF%99%E4%B8%AA%E4%B8%8D%E8%BE%A3%EF%BC%8C%E8%BF%98%E8%83%BD%E5%86%8D%E8%BE%A3%E7%82%B9%E5%90%97.webm>
- License / use policy: [CC BY 3.0](https://creativecommons.org/licenses/by/3.0/)
- Original time range: 10.35–28.80 seconds
- Local files: `samples/wangjerry_intro_zh2en/source.mp4` and `samples/wangjerry_intro_zh2en/translated.mp4`
- Source MP4 SHA-256: `681ed861e7f48b8ce539b08402e3a0b1ba8549cb7c9b96af9726ad205fdfc25b`
- Modifications: excerpt, downscaling, web encoding, soundtrack replaced with model-generated speech. No speech truncation or time stretching. Generated captions reproduce model text without manual correction.

## English → Chinese · Andrew Morgan: daily life & family

- Original title: T-60 Seconds with Andrew Morgan.webm
- Creator: NASA Johnson
- Original YouTube: <https://www.youtube.com/watch?v=Rww7zst5XhA>
- Source and video-specific rights record: <https://commons.wikimedia.org/wiki/File:T-60_Seconds_with_Andrew_Morgan.webm>
- License / use policy: [Public domain in the United States (NASA-created material); NASA media-use guidelines](https://www.nasa.gov/nasa-brand-center/images-and-media/)
- Original time range: 13.20–29.88 seconds
- Local files: `samples/morgan_family_en2zh/source.mp4` and `samples/morgan_family_en2zh/translated.mp4`
- Source MP4 SHA-256: `c861a726a043dfc2dae67349d899a2736f0b8f03b445c50f1bb248b265618b40`
- Modifications: excerpt, downscaling, web encoding, soundtrack replaced with model-generated speech. No speech truncation or time stretching. Generated captions reproduce model text without manual correction.

## Chinese → English · Wang Gang: preparing mapo tofu

- Original title: 厨师长教你川菜麻婆豆腐的正宗做法.webm
- Creator: 美食作家王刚 (Chef Wang Gang)
- Original YouTube: <https://www.youtube.com/watch?v=RU-0vjjvSiI>
- Source and video-specific rights record: <https://commons.wikimedia.org/wiki/File:%E5%8E%A8%E5%B8%88%E9%95%BF%E6%95%99%E4%BD%A0%E5%B7%9D%E8%8F%9C%E9%BA%BB%E5%A9%86%E8%B1%86%E8%85%90%E7%9A%84%E6%AD%A3%E5%AE%97%E5%81%9A%E6%B3%95.webm>
- License / use policy: [CC BY 3.0](https://creativecommons.org/licenses/by/3.0/)
- Original time range: 4.00–18.55 seconds
- Local files: `samples/mapo_preparation_zh2en/source.mp4` and `samples/mapo_preparation_zh2en/translated.mp4`
- Source MP4 SHA-256: `9196456a8978c99fd64751bb5084373d7e522f3f873b97c723b82a309145c492`
- Modifications: excerpt, downscaling, web encoding, soundtrack replaced with model-generated speech. No speech truncation or time stretching. Generated captions reproduce model text without manual correction.

## English → Chinese · Christina Koch: lessons from space

- Original title: Hear from Record-Breaking NASA Astronaut Christina Koch.webm
- Creator: NASA
- Original YouTube: <https://www.youtube.com/watch?v=UncCrD1cGb4>
- Source and video-specific rights record: <https://commons.wikimedia.org/wiki/File:Hear_from_Record-Breaking_NASA_Astronaut_Christina_Koch.webm>
- License / use policy: [Public domain in the United States (NASA-created material); NASA media-use guidelines](https://www.nasa.gov/nasa-brand-center/images-and-media/)
- Original time range: 586.78–616.20 seconds
- Local files: `samples/koch_lessons_en2zh/source.mp4` and `samples/koch_lessons_en2zh/translated.mp4`
- Source MP4 SHA-256: `dce6afa076c7b69d80084fdd32f1a0a097ef629f5d3e88352e7f301b5028ee06`
- Modifications: excerpt, downscaling, web encoding, soundtrack replaced with model-generated speech. No speech truncation or time stretching. Generated captions reproduce model text without manual correction.

NASA material is credited for provenance; DuraS2ST-generated audio and its accuracy are not NASA statements or NASA-approved translations. NASA source footage is public domain in the United States, subject to its media-use guidelines and applicable third-party rights.

## Henan Chinese → English: Wikitongues

- Title: *WIKITONGUES — Ying speaking Henan Chinese*
- Creator: Wikitongues / Ying Li
- Source page: <https://commons.wikimedia.org/wiki/File:WIKITONGUES-_Ying_speaking_Henan_Chinese.webm>
- Original video reference: YouTube `LZoRgfFdXXQ`
- License: [Creative Commons Attribution-ShareAlike 4.0 International](https://creativecommons.org/licenses/by-sa/4.0/)
- Local file: `source/henan_zh2en_01.mp4`
- Selected time range from the original: `00:00:15–00:01:00`
- SHA-256: `59ac16b991de27b6759d98b8afa293be6998b724904bb2f23c2261d62f813713`
- Attribution: Wikitongues / Ying Li, CC BY-SA 4.0.


## Guiyangese → English: Wikitongues

- Title: *WIKITONGUES — Changjiu and Chaofen speaking Guiyangese*
- Creator: Wikitongues / Brian Zhao
- Source page: <https://commons.wikimedia.org/wiki/File:WIKITONGUES-_Changjiu_and_Chaofen_speaking_Guiyangese.webm>
- Original video reference: YouTube `CbiVj4rV3dY`
- License: [Creative Commons Attribution-ShareAlike 4.0 International](https://creativecommons.org/licenses/by-sa/4.0/)
- Local file: `source/guiyangese_zh2en_01.mp4`
- Selected time range from the original: `00:00:15–00:01:00`
- SHA-256: `092af5c7d1bec5e39d3aa304c90d42e1c1b6d94ea8eb78f3902b23ec979c0e95`
- Attribution: Wikitongues / Brian Zhao, CC BY-SA 4.0.


## Processing note

New samples use sentence and pause boundaries checked against source transcripts, ASR and (where available) burned-in subtitles; Silero VAD was used to refine the mapo-tofu boundary. The released README inference functions and local checkpoint generate translated speech. Complete waveforms are placed on the original timeline with up to 0.36 seconds of observed start delay to prevent overlap (configured maximum: 0.40 seconds). No lip-sync alignment is claimed.

The retained Wikitongues examples use the earlier Silero VAD workflow. Their `translated_pure.mp4` has generated speech; `translated_mix.mp4` includes a lowered/separated background track. These adapted Wikitongues clips remain under CC BY-SA 4.0. They were not re-evaluated in this refresh.
