Does anyone know of any open source simlish/gibberish speech generators, or games that implement these?
I'm referring to where instead of recorded voice, you have gibberish audio played while speech text is shown. The audio can even dynamically fit the text to varying degrees. A very popular approach in the 90's is where a single sound effect is played repeatedly, with possible pitch and speed variations, in sync with text. This was used a lot in classic JRPGs, but also many other games like GTA 1/2. Here's an example from Shining Force II:
The approach is very simple; it assigns a sound for each individual letter and plays them with significant overlap between the letters, as described in this reddit thread: https://www.reddit.com/r/gamedev/commen ... e_blarble/ It mostly works even with English; I imagine it would work better with languages with more consistent phonemes. The downside is that it has a distinctive stutter-like sound which may not fit certain themes/moods.
BTW: What method was used for this one: https://www.youtube.com/watch?v=3RotjK246l4 Approach #2? Is the length of the patterns usually set manually or can that be handled by software to support multiple text languages?
That sounds more like an advanced version of #1 - it's still using a single sound effect, but rather than a synthetic one like JRPGs, it's using a voice sample. The hardware randomises the speed and pitch. It sounds like the same one used in Banjo-Kazooie:
This isn't surprising; both games were made around the same time so the motivation would be the same: game consoles could handle audio samples but had limited space to store large amounts of voice acted content, so this was one way to save space (and money obviously). The guy who made this, veteran game composer Grant Kirkhope talks more about it here: http://www.gameinformer.com/b/features/ ... areer.aspx
P.S. are youtube embeds broken or am I doing it wrong?
This is a pretty interesting topic as it can really help story driven games that really have problems with silent dialogs. The animalese.js example sounds a bit too close to english though and unless you want to go for something intentionally funny it is probably not a good choice.
“You never change things by fighting the existing reality. To change something, build a new model that makes the existing model obsolete” - Buckminster Fuller
I found a related open source project: https://github.com/mnater/hyphenator which is an algorithm that inserts hyphens at syllable boundaries, used in TeX, but you could in theory also use it to detect syllables as part of a gibberish speech generator.
I found the problem with youtube embeds; my browser is blocking them because the forums are HTTPS but the embed URL is HTTP. If I change the embed to HTTPS they show up again.
I've been thinking about this topic for some time... Each game would require a bit different approach to gibberish generation. E.g. one needs some funny gibberish, another needs serious one... One game would be "fine" with just one sound pitch-shifted, while another won't be fine even with "letter-to-letter" algorithm and would require processing diphthongs/syllables or something like that. Some games even just used a "typewriter" sound to represent speech. https://youtu.be/H_7QZ6Q-sHs?t=56s Or just some random (sometimes annoying) sound https://www.youtube.com/watch?v=oPyzaKoGjks Actually my favourite is https://www.youtube.com/watch?v=CxQTmCA_KA0
I would record 3 or 4 gibberish noises (or however many you want) which should be very short, maybe half a second each (you'll have to experiment) then just play them randomly using the length of the text to decide how long a chain of random gibberish sounds you want to put together.