Creative Communities of the World Forums

The peer to peer support community for media production professionals.

Activity Forums Creative Community Conversations Generate subtitles from script instead of hoping for “AI” to “get it” from audio

  • Generate subtitles from script instead of hoping for “AI” to “get it” from audio

    Posted by Jacob Giacometti on March 15, 2025 at 9:18 am

    Salut !

    As professional film makers, I’m sure many of us start off the subtitling or captions process WITH the actual official proper clean script.

    As of now, all options I find involve some kind of approximative “AI” analysis of the audio, resulting in very approximative text, with wrong pronunciations, names, punctuations etc…

    I would have thought that by order of priority, before addressing this more complex method, the more basic method would be covered : instead of asking the robot to understand the audio, we give the text to the robot. His only job is to now align it with the audio, instead of having to figure it out.

    Has anyone been able to develop a workflow, or even better a software that does just that ?
    I am aware that Youtube, vs all other paid editing software, the irony, does offer that feature, but it will auto-align a text only in certain languages. It is also not a super convenient workflow having to go back and forth from editing software to youtube to do it.

    I’m on Mac, using Resolve, but I’m open to other softwares if it works !

    Thanks for your insights ! 🙏🏼


    Michael
    Current work : ANINAGIN, the Aliens from Earth (self produced)
    https://www.youtube.com/channel/UCC71RypUFAfLnXEVIzw-llA

    Mads Nybo jørgensen replied 1 year ago 2 Members · 3 Replies
  • 3 Replies
  • Jacob Giacometti

    March 15, 2025 at 9:56 am

    PS : The proper technical term seems to be called ” Forced Alignment” 🫨

  • Mads Nybo jørgensen

    May 31, 2025 at 6:01 pm

    Hey Jacob,

    Sorry, I stored this away in a tab amongst 1000s on FireFox, and today as destiny would have it, it was presented to me.

    You make some interesting points, and raise good questions.

    It is only in scripted productions (mainly drama) that you get a script. Some larger organisations create scripts from their unscripted shows at the end of the Post production. But to match it to the spoken word might be difficult – although, I’m sure someone could write a plug-in for that.

    Although I have not worked with it, yet (have got the licence), Avid Scriptsync might do what you are looking for:
    https://www.avid.com/products/media-composer-scriptsync-option

    On a wider note about speech to subtitles, I am not sure that A.I. is the right wording to use in the context of what is effectively “Machine Learning”?

    However, if you want to “reverse” the process. Where the SOT (Sound On Tape) is transcribed, then fixed for errors, and turned into a “new” voice that is inserted into the right places on the edit. And, spaced out to match lips. Then, you’ll get rid of the original speaker, but use A.I. to mimic their voice. Then you are looking at something very different, that might not be received well by the person with the actual voice.

    Only a week ago Epic games, with their new A.I. voice bots, walked into controversy on the subject of replacing the original voice actor with A.I. (Although some might argue that it will take a shovel and some amount of raising the ead from the Grave, to get Darth Vader’s original voice.

    https://www.bbc.co.uk/news/articles/cn844l3eznqo

    Just an opinion.

    Atb
    Mads

    PS: Love your AninagiN trailer, will watch the movie!

We use anonymous cookies to give you the best experience we can.
Our Privacy policy | GDPR Policy