Speech to Text does not transcript Clips (videocodec-typ: MP4/MOV H.264 10 bit 4:2:2)

Report · May 23, 2023

Speech to Text does not transcript clips ( MP4/MOV H.264 10 bit 4:2:2) from my Panasonic Lumix S5 II with following specifications:

Typ: MPEG-Film Dateigröße: 26,35 GB Bildgröße: 3840 x 2160 Framerate: 25,00 Audioformat der Quelle: 48000 Hz - 24 Bit - Mono Audioformat des Projekts: 48000 Hz - 32-Bit-Floating-Point - Mono Gesamtdauer: 00:24:31:17 Pixel-Seitenverhältnis: 1,0 Alpha: Keine Farbraum: Rec. 709 Farbraum überschreiben: Aus Eingabe-LUT: Keine Videocodec-Typ: MP4/MOV H.264 10 bit 4:2:2

When I transcode the Clip to ProRes, transcription of the clip is possible.

I can also transcribe a sequence with the original clip ( MP4/MOV H.264 10 bit 4:2:2) inside.

Is there any idea how to make the transcripption of the original clip possible?

Thank you, Stefan

PP Version 23.4.0 (Built 56)

Apple 13.3.1; M1 Max 32GPU

Report · May 23, 2023

When you drag the clip to the "new" icon (and create a sequence with the same settings), what do you see for audio tracks?

Source media transcription will only transcribe the first audio track. There is a workaround, but let's see what you have first.

Stan

Report · May 23, 2023

Hi Stan,

thank you for your quick reply. There is 4 audio Channels and Channel 1 is empty. (Channel 2 is a Camera backup Mic, Channel 3 is Lavallier Person A, Channel 4 is Lavallier Person B).

Stefan

Report · May 24, 2023

Yeah, it's probably because Channel 1 is empty. Can you try modifying the audio channels of the source media so that channel 1 corresponds to the channel you want transcribed?

Right click on source media in the bin > Modify > Audio Channels, re-assign the channels and see if it works.

Report · May 24, 2023

Unfortunately reassigning the channels is not working.

Report · May 24, 2023

That's a shame, it should have a window similar to when you transcribe a sequence, where you can tell it with channel to look at. We should create a feature request for this.

Report · May 25, 2023

I agree that reassigning channels does not work. It appears to look at track 1 in the actual file, not in virtual changes in PR. I really don't know.

I call it "track 1" because I cannot tell what it is doing regarding "channels." One user in another thread had a file with one stereo track and one person in the left channel and another in the right. It transcribed with both speakers indicated.

In another thread, a user had a file with track 2 being microphone and track 3 being another dialogue input. I suggested the method of using a static transcript of the clip in a sequence and repeated that twice. The editing sequence became the video, track one system audio, track two nest of original track 2 with a static transcript that behaves as a source transcript, and similar track 3.

I also played with extracting audio inside PR, but I think that has a greater chance of running in to sync problems.

Stan