Wav output duration different from source webm

[le-ro-la]

I'm Leonardo from Iguale, a Brazilian accessibility company. We produce audio description, sign language and subtitles for films.
The encoder has been a great help when we generate the sign language for DCP.

I would like to understand a question that has caused us doubts:

We generate the files in mp4. From these, we generate the Webm. Then we use your encoder to create the .wav files in Vp9. However, wav outputs have a different duration than webm or even mp4. Please note the following list:

Duration and TC of the Reels and Sign Languages.Mp4:
R1 20:55 (01:20:54:15)
R2 16:16 (02:16:15:20)
R3 14:26 (03:14:25:18)
R4 20:06 (04:20:05:14)
R5 33:40 (05:33:40:11)
R6 00:17 (06:00:16:13)

Webm duration:
R1 20:54
R2 16:15
R3 14:25
R4 20:05
R5 33:40
R6 00:16

VP9 wav duration
R1 20:56
R2 16:16
R3 14:26
R4 20:06
R5 33:42
R6 00:18

Is this normal? Or are we experiencing a bug in the middle of the process?
We were asked by the company that finalizes the DCPs about this difference in the length of the files.

Could someone please give us some light on this?

We thank you in advance,

[xultz]

I'm not working anymore with this topic, but I suppose that the difference is because the WAV-VP9 protocol embeds the video in 2 seconds lenght of wav audio. If, for example, the original video has 36.1 seconds, the resulting wav will have 38 seconds. It makes some sense, by looking that all the VP9 duration you produced has even values. [Now in portuguese, to be clearer] Eu não trabalho mais com este assunto há bastante tempo, mas eu desconfio que o resultado que obteve é porque o protocolo de conversão de vídeo para WAV-VP9 empacota o vídeo em pedaços de 2 segundos de áudio. Se, por exemplo, teu vídeo original tiver 36,1 segundos de duração, o WAV terá 38 segundos (aqueles 0,1 segundos do final entram num pacote de 2 segundos). Essa explicação faz sentido, se você observar que teus resultados todos têm valores pares (ou seja, é possível dividir aquele áudio em pedaços iguais de 2 segundos). Se essa diferença de comprimento pode impactar no DCP, em teoria o WAV produzido poderia ter seu final cortado de modo a deixar com o comprimento exato do vídeo do Reel. Com certeza absoluta, esse corte não traria nenhum problema de decodificação no ProAccess da Riole (digo isso, porque essa decodificação fui eu que fiz), acredito que no Dolby também não tenha problema (mas não posso afirmar). Se quiser aprofundar essa conversa, pode me mandar um email. Abraço!

…

On Tue, Mar 5, 2024 at 11:22 AM le-ro-la ***@***.***> wrote: I'm Leonardo from Iguale, a Brazilian accessibility company. We produce audio description, sign language and subtitles for films. The encoder has been a great help when we generate the sign language for DCP. I would like to understand a question that has caused us doubts: We generate the files in mp4. From these, we generate the Webm. Then we use your encoder to create the .wav files in Vp9. However, wav outputs have a different duration than webm or even mp4. Please note the following list: Duration and TC of the Reels and Sign Languages.Mp4: R1 20:55 (01:20:54:15) R2 16:16 (02:16:15:20) R3 14:26 (03:14:25:18) R4 20:06 (04:20:05:14) R5 33:40 (05:33:40:11) R6 00:17 (06:00:16:13) Webm duration: R1 20:54 R2 16:15 R3 14:25 R4 20:05 R5 33:40 R6 00:16 VP9 wav duration R1 20:56 R2 16:16 R3 14:26 R4 20:06 R5 33:42 R6 00:18 Is this normal? Or are we experiencing a bug in the middle of the process? We were asked by the company that finalizes the DCPs about this difference in the length of the files. Could someone please give us some light on this? We thank you in advance, — Reply to this email directly, view it on GitHub <#5>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEVLITV26WGHJS62JDQFDILYWXILNAVCNFSM6AAAAABEHHSIBKVHI2DSMVQWIX3LMV43ASLTON2WKOZSGE3DSMZWGI2DEMI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[le-ro-la]

Hi Christian, Thank you very much for the clarification. It makes total sense! It's cool that you worked with Riole, we've already had contact with them on other projects. Just one last question: So can I cut the wavs in any audio program? Best *___________________________________________* *Leo Rossi* *Locutor - Dublador - Audiodescritor - Voice Over Artist* *www,leorossi.com.br <http://www.leorossi.com.br>* https://www.instagram.com/lerolero_la/ *Fone/WhatsApp - + 55 19 99788 8581**PLEASE NOTE: By receiving this email, you guarantee that you will not use the material narrated by me (Leo Rossi), or part of it, as a subsidy for voice cloning, resulting in the same or a third. **Failure to comply with this agreement will result in legal proceedings.* Em ter., 5 de mar. de 2024 às 11:37, Christian Schultz < ***@***.***> escreveu:

…

I'm not working anymore with this topic, but I suppose that the difference is because the WAV-VP9 protocol embeds the video in 2 seconds lenght of wav audio. If, for example, the original video has 36.1 seconds, the resulting wav will have 38 seconds. It makes some sense, by looking that all the VP9 duration you produced has even values. [Now in portuguese, to be clearer] Eu não trabalho mais com este assunto há bastante tempo, mas eu desconfio que o resultado que obteve é porque o protocolo de conversão de vídeo para WAV-VP9 empacota o vídeo em pedaços de 2 segundos de áudio. Se, por exemplo, teu vídeo original tiver 36,1 segundos de duração, o WAV terá 38 segundos (aqueles 0,1 segundos do final entram num pacote de 2 segundos). Essa explicação faz sentido, se você observar que teus resultados todos têm valores pares (ou seja, é possível dividir aquele áudio em pedaços iguais de 2 segundos). Se essa diferença de comprimento pode impactar no DCP, em teoria o WAV produzido poderia ter seu final cortado de modo a deixar com o comprimento exato do vídeo do Reel. Com certeza absoluta, esse corte não traria nenhum problema de decodificação no ProAccess da Riole (digo isso, porque essa decodificação fui eu que fiz), acredito que no Dolby também não tenha problema (mas não posso afirmar). Se quiser aprofundar essa conversa, pode me mandar um email. Abraço! On Tue, Mar 5, 2024 at 11:22 AM le-ro-la ***@***.***> wrote: > I'm Leonardo from Iguale, a Brazilian accessibility company. We produce > audio description, sign language and subtitles for films. > The encoder has been a great help when we generate the sign language for > DCP. > > I would like to understand a question that has caused us doubts: > > We generate the files in mp4. From these, we generate the Webm. Then we > use your encoder to create the .wav files in Vp9. However, wav outputs have > a different duration than webm or even mp4. Please note the following list: > > Duration and TC of the Reels and Sign Languages.Mp4: > R1 20:55 (01:20:54:15) > R2 16:16 (02:16:15:20) > R3 14:26 (03:14:25:18) > R4 20:06 (04:20:05:14) > R5 33:40 (05:33:40:11) > R6 00:17 (06:00:16:13) > > Webm duration: > R1 20:54 > R2 16:15 > R3 14:25 > R4 20:05 > R5 33:40 > R6 00:16 > > VP9 wav duration > R1 20:56 > R2 16:16 > R3 14:26 > R4 20:06 > R5 33:42 > R6 00:18 > > Is this normal? Or are we experiencing a bug in the middle of the process? > We were asked by the company that finalizes the DCPs about this difference > in the length of the files. > > Could someone please give us some light on this? > > We thank you in advance, > > — > Reply to this email directly, view it on GitHub > <#5>, or > unsubscribe > < https://github.com/notifications/unsubscribe-auth/AEVLITV26WGHJS62JDQFDILYWXILNAVCNFSM6AAAAABEHHSIBKVHI2DSMVQWIX3LMV43ASLTON2WKOZSGE3DSMZWGI2DEMI> > . > You are receiving this because you are subscribed to this thread.Message > ID: ***@***.***> > — Reply to this email directly, view it on GitHub <#5 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BFPPZRNE4W6UZ6LSFA6WG4TYWXKCNAVCNFSM6AAAAABEHHSIBKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNZYHEYTONRVG4> . You are receiving this because you authored the thread.Message ID: ***@***.***>