Workflow for Video Transcribing

[greymalkin]

All,

This isn't technically subtitle related but this is the closest topic I could find to the subject .

I'm looking to transcribe roughly 30 hour long videos that are speech only (think talking head/no other audio).

I'm looking for advice on your best workflows you have used..is any speech to text really THAT MUCH better than the others? I imagine this will be a scenario where we play the video, watch the speech to text and intervene when it inevitably gets things wrong.

Free software is always preferred but if there's something that gives noticeably better results for a price (one time fee preferably, not interested in subscription or something where you send it off to another person to correct...we will correct)

Thanks in advance to anyone with experience in this area who takes the time to reply!

[VoodooFX]

...is any speech to text really THAT MUCH better than the others?

Whisper model is the best in general, I recommend Faster-Whisper, both implementations you can find in my signature.

[Subtitles]

All,

This isn't technically subtitle related but this is the closest topic I could find to the subject .

I'm looking to transcribe roughly 30 hour long videos that are speech only (think talking head/no other audio).

I'm looking for advice on your best workflows you have used..is any speech to text really THAT MUCH better than the others? I imagine this will be a scenario where we play the video, watch the speech to text and intervene when it inevitably gets things wrong.

Free software is always preferred but if there's something that gives noticeably better results for a price (one time fee preferably, not interested in subscription or something where you send it off to another person to correct...we will correct)

Thanks in advance to anyone with experience in this area who takes the time to reply!

A lot will depend on your computer specifications.
To get accurate transcription you will need a decent spec with a GPU 12GB VRAM but if you don't have this then 8GB VRAM will give you a decent accuracy.

Subtitle Edit has several free Audio to Text modules so you can try that to start with. Just try a about 15 minutes audio clip and see how good it is for you.
I wouldn't recommend transcribing the whole 30 hours in one go. Split into logical parts of maybe 2 hours each and correct the errors if you find any.

I use Whisper AI and I am happy with it. It is a Command Line (cmd) software.
https://github.com/openai/whisper

All this is free but you can try subscription options that have audio to text features.
I think Adobe Premiere has it so perhaps you can try one month subscription for example.
Other brands like Davici Resolve and Vegas Pro 365 have audio to text features but from my experience the trial versions don't include audio to text.
The Vegas Pro 365 has a limit on how many hours of transcription a YEAR so that is useless, unless you pay more to get more hours. Ridiculous IMHO

[Bernix]

Or wait until VLC comes out with its version of AI subtitles. It shouldn't take long.