With the new "Autovoice" function in Storykit, we have made it possible for you to create a basic voice-over automatically from your storyboard. This voice-over can be printed to your soundtrack, and you can also download it for other uses.
The use of "Autovoice" helps you in situations where you need to have some accessibility functionality. It should not be a substitute for a voice actor read speaker text when you are in need of such, but could be used for giving sight or reading impaired viewers a way to follow the script even on platforms where a screen reader may not be functional or when you can't use a sidecar script file for that screen reader.
The technology behind this is Google Wavenet, a state-of-the-art speech synthesis solution that often yields a better readout than screen reader technology.
Re-calculation of the storyboard
It's good to know how this function works to make use of it and prepare your storyboards and scripts in the best way. The "Autovoice" generation will look at each slide and read them in full. So the general rule is that text that is visible on-screen is being read when visible.
Since duration is set per slide, this is also where the duration calculation will take place. So when creating a voice-over version of a storyboard, which is a specific function, we will look through your storyboard with our Coen AI-based duration algorithm and try and make sure that you do not have too short slides for the voice-over to work. When we find those cases, we will change them for you in the new storyboard copy. So the voice-over storyboard can get a longer duration, we never shorten slide duration – since that is a creative decision on your part to have them longer.
So, in short: We try our very best to make sure that your script can be read by the "Autovoice" in time and in full.
Instructions for use
You create a voice-over version when you are done with your storyboard; it makes no sense to do it before since the voice-over is generated from your script. Before doing this – please read through the best practice suggestions below; there is a lot you can do to make the output a lot better.
Creating the voice-over version
- From the Overview page, access the more-menu (...) for the storyboard you want to use.
- Choose "Create Voice-over Version."
- In the new modal, you can give your new storyboard a name; then, you choose the script's language before clicking "Create Voice-over."
- This will now create a new copy of your storyboard, with any duration changes that are deemed necessary.
- You will now see a microphone icon next to the new storyboard. This is what shows you that this is a voice-over storyboard.
- To render your storyboard with the new voice-over, open the storyboard and generate it as you would typically do. If you have selected background music, or you do this at this moment, the voice-over and background music will mix. You can also leave it without music to get only the voice-over.
- Note: When having a voice-over – the music level gets lowered by -12db to not interfere with the spoken voice-over as much. This makes the speaker voice more audible and better suited to its purpose.
Alternate text for voiceover
You can let your voiceover read something other than the written script using the "Alt Description" in the right panel. When you input text into that text field, the Autovoice function will use this text instead of the slide text.
Removing slide text from voiceover
You can also remove all text for a slide (slide text and alternate text) by using the "Ignore all text" checkbox in the "Alt Description" part of the right panel. Doing so will ignore all text in the voiceover generation.
Downloading the voice-over audio file
If you need it, the "clean" voice-over audio is available for you to download. This is how you do it.
- This needs to have been generated first for this option to appear.
- From within the storyboard, open "Music."
- You find your voice-over file at the very top of the list (a .wav file). Clicking the file name next to "Download voice-over" will download that file to your computer.
Best practice – do this to get good results.
You can use quite a few tricks to yield a better and more coherent voice-over. Do as many of these as you can – before creating the voice-over version, and you should be able to get something that is very functional.
- Avoid words that are too complex: As in all written text, make the extra effort to see if some words and phrases are a bit simpler. When creating the voice-over, the reason for doing this is that the speech synthesis technology often is a lot better reading out simpler words.
- For compound-word languages – play with a hyphen: In languages that use many compound words – like Swedish – the synthesis engine sometimes will pronounce those words a bit strange. When having to use rather exotic compounds, you can try to add a hyphen between the word parts just when generating. This can sometimes force the synthesis engine to pronounce the words more clearly.
- Don't cut up sentences: Since the synthesis works on a slide-per-slide basis, cutting up sentences over multiple slides works so-so. The speech synthesis will have no connection between the different parts of the sentence, making the reading a bit strange. So try and have complete sentences for every slide if you can.
- Avoid abbreviations and parenthesis as much as possible: Something that can trip up the timing is slides with many abbreviations. So try and keep those to a minimum – this is also good practice for readability in general. Parenthesis is also something that trips the speech engine up a bit. And read further on for a twist on this theme. Also, things like # – which will read out as "hashtag" in many languages, can make the timing a bit uncertain.
- Write company names as words: Many companies demand that their name is spelled in only upper case. If the company name is meant to be read out – like "IKEA" – you want to write it as "Ikea." This means that the synthesis engine views it as a word and reads it. Otherwise, there is a risk that it will read it letter by letter. There are many cases where that is preferred, and then you can leave the name in all upper case. But if you want to make sure it gets read out – change it to a mixed case before generating.
- Try to avoid language mixing: You set your voice-over to one specific language, meaning that it will read everything in that synthesis engine. The different language versions can read things in other languages, but it's not great. So if you have a situation with many borrowed words – widespread with English titles, for example – try and rephrase something in your source language.
What does not work
- The timing algorithm is always subject to change and refinement.
- For now, "Autovoice" only supports English, German, Swedish, Danish, Norwegian, French, Italian and Spanish. More languages to come.