AI assisted transcription and subtitle creation

All hardware and software related discussion topics here. Advice, discussion and opinions on either topic are welcome.
User avatar
Nasuada
Posts: 645
Likes:
Joined: Sat Jun 06, 2020 1:05 pm

Re: AI assisted transcription and subtitle creation

Post by Nasuada »   2 likes

A dream slowly come truth. The only problem is... If we start to trust to much the ki translation and do not let check them by a native language talking person, the KI can say us anything, we trust it and we do not know it's really the right story. If a KI change something or made some important parts wrong, this can change anything around a story.

I'm in a good hope that this will not hapen here on FLM, because there are some movie maniacs here from different countries. But my personal opinion at this point can be a special marking for KI translated stuff.
[Image]
User avatar
ghost
Site Admin
Posts: 8460
Likes:
Joined: Sun Mar 07, 2004 1:00 am

Re: AI assisted transcription and subtitle creation

Post by ghost »   1 likes

Whisper in action using Google drive:
[Image]

This the Danish short film Et liv.
User avatar
ghost
Site Admin
Posts: 8460
Likes:
Joined: Sun Mar 07, 2004 1:00 am

Re: AI assisted transcription and subtitle creation

Post by ghost »   1 likes

French to English:
[Image]

That's Va, petite!
User avatar
Night457
Global Moderator
Posts: 5221
Likes:
Joined: Sat Dec 28, 2019 3:44 pm

Re: AI assisted transcription and subtitle creation

Post by Night457 »   1 likes

Nasuada wrote: Mon Mar 06, 2023 8:53 pm A dream slowly come truth. The only problem is... If we start to trust to much the ki translation and do not let check them by a native language talking person, the KI can say us anything, we trust it and we do not know it's really the right story
I understand your worry because it is now widely reported that ChatGPT and other AI chatbots are relentless LIARS. And there is no replacing human translators in explaining, um, human language. I don't have much confidence in proper translations of slang expressions, because in some cases I have explored THE ENTIRE INTERNET to find an answer to the question "What the heck does THIS Swedish expression mean?" And the answer comes in Swedish blogs: "The Northern Swedes know but they will not tell the rest of us in the south." So if Swedes do not tell other Swedes then a damn computer is not going to figure it out.

HOWEVER ... don't trust the machine overlords in real-world scenarios, but make use of them in the realm of fiction, such as translating subtitles for a movie. It is just entertainment, after all. I seriously doubt that a cute little Swedish girl is going to volunteer to work for free translating Astrid Lindgren movies for me! Even with a machine-translation as a start (still computer Artificial Intelligence involved there), I spend many hours afterward exploring multiple dictionary sources online. Yet the Swedish just does not stick in my head. (The old dog, new tricks story.) So if I can make the work take a little less time with the help of a lying robot sociopath, why not? I just have to double-check what seems unlikely to my bullshit-detector.
No 9000 computer has ever made a mistake or distorted information. We are all, by any practical definition of the words, foolproof and incapable of error.
...
I know I've made some very poor decisions recently, but I can give you my complete assurance that my work will be back to normal. I've still got the greatest enthusiasm and confidence in the mission. And I want to help you.
HAL.jpg
You do not have the required permissions to view the files attached to this post.
User avatar
bonmrmjp
Posts: 36
Likes:
Joined: Mon Jun 27, 2022 4:45 am

Re: AI assisted transcription and subtitle creation

Post by bonmrmjp »   1 likes

I've been experimenting with whisper as well, and I'm very impressed. For example, I tried transcribing " Der Richter und das Mädchen", then using Google translate (in Subtitle Edit) to change it to English, and it worked impressively well. Not perfectly, but an order of magnitude better than google speech to text. It works well enough to follow the movie. The text seemed to be at least 95% accurate. In my case, I've got a year of high school German, so I at least know the basics, which helped cleaning it up. However, its not that good at matching speaking times, so there is considerable time that it takes adjusting start & end times in Subtitle edit. But its definitely an option for movies without any captions.

Another option I've done is to get a native speaker to transcribe the video using Fiverr.com. That's how we got the captions for "Ståtrold". That short had little dialog, so it was only $5 for the transcription.
barracute
Posts: 19
Likes:
Joined: Thu Jun 18, 2020 2:14 pm

Re: AI assisted transcription and subtitle creation

Post by barracute »   1 likes

I'm using whisper but cant get it to trans a full clip of 14mins? large V1 gives 11min large v2 4mins and just large 10mins? how can you get it to do the full clip????
User avatar
DreamScape
Posts: 291
Likes:
Joined: Tue Sep 15, 2020 7:06 am

Re: AI assisted transcription and subtitle creation

Post by DreamScape »   0 likes

ghost, are you not getting a crazy amount of repeats? Or have you not tried a feature length movie yet?

[Image]
User avatar
ghost
Site Admin
Posts: 8460
Likes:
Joined: Sun Mar 07, 2004 1:00 am

Re: AI assisted transcription and subtitle creation

Post by ghost »   0 likes

DreamScape wrote: Wed Mar 08, 2023 6:04 pm ghost, are you not getting a crazy amount of repeats? Or have you not tried a feature length movie yet?

[Image]

Yes, I got also some repeats, but not as crazy as you did. :o

I figured out, that cleaning up the audio before helps to get almost rid of them. https://www.lalal.ai/voice-cleaner/
David32441
Posts: 799
Likes:
Joined: Thu Jul 22, 2021 2:48 am

Re: AI assisted transcription and subtitle creation

Post by David32441 »   0 likes

Wow, this looks really good, and the Youtube video makes it look really easy!
User avatar
pillowbaker
Posts: 2111
Likes:
Joined: Mon Mar 07, 2022 4:05 am

Re: AI assisted transcription and subtitle creation

Post by pillowbaker »   0 likes

ghost, do you suggest using that voice-cleaner software every time before using whisper AI on it?
Post Reply