Forum Discussion

ThierryEMMANUEL's avatar
ThierryEMMANUEL
Community Member
27 days ago

Problem with automatically generated subtitles ?

Hi everyone.
I've detected a problem with the voices generated by Storyline's AI and the subtitles that are automatically generated at the same time.
I copy and paste a prompt and check the “Generate subtitles” box. Everything is perfect. Of course, the subtitles are a mess, as usual. It takes me a long time to cut and arrange them perfectly.
Then I decide to change (for example) a single word in my prompt. To preserve my work on the subtitles, I uncheck the “Generate subtitles” box. As a result, the subtitles are still completely recreated, once again in complete disarray.
Is there something I'm not understanding (between “Generate subtitles” and NOT “generate subtitles”)?
Should I do something else (other than exporting my subtitles and reimporting them)? This seems like a bug, doesn't it? Has anyone else noticed this? (I haven't found any previous posts on the subject). It seems to me that it worked logically and normally in a previous version: when I uncheck the “Generate” box, the subtitles that have already been created are not damaged. Am I wrong?

Have you noticed the same thing? Any ideas or tips?

Thanks

Thierry

6 Replies

  • It seems I was mistaken when I said “it worked logically and normally in a previous version: when I uncheck the ‘Generate’ box, the subtitles that have already been created are not damaged.” This is the case with neural or standard voices (which are absolutely unusable in French, too ugly, by the way), but it does not apply to AI-generated voices. So, back to my questions. Does anyone else have a problem with these subtitles?

  • Hi ThierryEMMANUEL​!

    Sorry to hear you're having trouble with AI-Text-to-Speech Captions in Storyline 360!

    We're tracking a bug in Storyline which causes captions to show in the caption editor, even if the author opts out of generating them. I've linked this discussion to the bug report so we can notify you as soon as a fix is ready!

    Also, you mentioned captions appearing messy and requiring rearrangement. This shouldn't be happening. However, we are tracking a separate Storyline bug which causes behavior similar to what you've reported. When "Generate Closed Captions" is enabled, the captions do not align with the natural pauses in AI-generated audio or Text-to-Speech. Do you mind sharing a copy of your affected .story file so we can confirm if you're being affected by the same issue?

    If you'd prefer to share your file privately, please do so through a support case.

    Looking forward to hearing from you!

    • ThierryEMMANUEL's avatar
      ThierryEMMANUEL
      Community Member

      Hello StevenBenassi​ 

      Thank you for your reply. I am happy to send you an extract from one of my .story files (a few screens) that I have annotated for clarity. But I am almost certain that it is not my files that have a problem, but rather Storyline in this regard.
      I can confirm that the subtitles are never “aligned with the natural pauses in AI-generated audio or Text-to-Speech.”


      What I discovered today is that if the “Generate subtitles” box is unchecked, then the subtitles, although visible in the subtitle editor, do not appear on the screen. OK. But if I want to make them appear after all, by checking the box, I get disorganized subtitles every time.


      I know there are workarounds. Export/import neatly arranged subtitles.vtt. Cut the voices into several pieces so that you only have to rework ONE sentence if you need to change ONE word (but we know that the quality is better when the voice says an entire paragraph, not just one sentence.)
      But there's something impractical about it.

      The expected logical behavior would be (in my opinion): 1) Subtitles aligned with natural pauses in the voice. 2) Automatic creation and modification of subtitles SUSPENDED when the “Generate subtitles” box is unchecked. 3) Manual arrangement of subtitles 
      RETAINED if the box is unchecked 4) Differentiation between the two functions: “Generate subtitles” and “Display subtitles.”

      Please don't forget: why do I insist so much on being able to keep manual changes? In French (and many other languages!), subtitles need to provide additional nuance. For example: the voice + subtitles say “Ready?”. In French: “Prêt ?” But if you're addressing a woman: “Prête ?” (with an e). And to several people: “Prêts ?” (with an s). So, to address everyone, and to be quick, I'll write “Prêt(e)(s) ?” in the subtitles. This is allowed. But I would lose this work every time I modified the voice prompt. There are dozens of examples in French. Hundreds of examples if you count other languages.
      Thank you in advance.

      • ThierryEMMANUEL's avatar
        ThierryEMMANUEL
        Community Member

        A new thought occurred to me this afternoon, dear StevenBenassi​.

        On my latest project, my client asked me to change the AI voice on the finished module, without changing the text. I would have liked to be able to keep the precise re-cutting of all my subtitles by changing only the voice!


        And one last thing before I go for a nap: as you can see in my file, my prompts have lots of line breaks to force the voice to pause where I want it to, since SSML tags don't work with AI voices (am I right?). This may also prevent the subtitles from aligning properly with the “natural” pauses in the AI voice. 

  • Just two quick responses to continue providing feedback, dear StevenBenassi​ .
    I did a quick test. Indeed, if I create an audio file with BRIAN's AI voice (for example), the subtitles are a mess and unusable without patient reworking. If I switch to the GRACE voice, the subtitles are indeed unchanged, you're right! But they're still just a identical mess. 😅 What's the point? My problem is that if I had reworked them, I would have wasted my time again.

    As for classic TTS, the standard voices (in French! Is it the same in English?) are so ugly that I can no longer offer them to my clients, who are already used to much more pleasant, almost human voices in Storyline or elsewhere. And neuronal voices are almost always criticized for their mechanical tone. You yourselves have made these two options obsolete with the AI option. I have contracted it almost exclusively for AI voice creation (with the problems described above), not for image creation, never for content creation, given my particular situation: 100% storyliner, 0% instructional designer. Et voilà.

    But you already know the problems you have to solve. So, I wish you the best of luck with that. Kind regards.