Does a computerized speech module already exist?

Hello,

I searched the library but couldn’t find a module that makes retro-style synthesized vocal sounds - similar to a Speak & Spell. I’m considering building one but I don’t know if it’s been done before. Anyone seen anything like this yet?

Cheers
Bret

Nysthi has the SAM module. It seems to have the downside that the parameter controls don’t work while it is speaking the text and changes only take effect when it restarts speaking.

There’s a vocal synthesis mode in the Audible Instruments Macro Oscillator 2, but you can’t use your own text as input, there’s just a selection of predefined words it will do.

2 Likes

Thank @Xenakios. Well… I probably wouldn’t bring much new to the table, so I suppose I’ll move on to other crazy ideas. :slight_smile:

what about a retro style 8bit sample player? i always liked the way midines’ sample bank sounded.

or maybe a module based on lsdj’s 59 phonemes or even the wave channel?

One downside of SAM is that it only says one line at a time, sort of hard to use if you want it to say a whole songs worth of lyrics. I use it with Stoermelder 8face2 but that only gives me 16 lines and makes things a little squrrelly to set up.

Just being able to load a list of lines of dialogue would be really useful but being able to change the pronounciation so you could get things to scan would be best.

OK, maybe I’ll look deeper into it. My first idea was to create a simple syllable generator of sorts. I had good luck before with this type of thing (https://github.com/clone45/EquationComposer/blob/master/experimental/ModuleVocalizer.cpp), but it’s been so long that I don’t have links to the original code which it’s based on. @persy I’ll take a look at those suggestions. I also have a big project in the works that I may focus on for a while. I’ve got to keep my mind occupied until I can go out fishing again. Ha ha ha.

1 Like

I’m dreaming about some vocoding Kraftwerk like voice in VCV Rack) If you’re gonna do some module in this vein (or not) there is a suggestion from me: each phrase has it’s own trigger, or like RUN (hold as gate or just on as trigger) and PAUSE triggers, which would be a great and smaller solution to make a whole song without so much mystery things)

you can sequence sentences in SAM just edit the “sam.dat” file in RES

1 Like

@VCVRackIdeas What if it worked like refrigerator magnets? But instead of placing them on your refrigerator (which has no concept of rhythm), instead you place the virtual magnets on a timeline, like this?

image

I had a little whiskey, so that might be very difficult to understand. It would play the top row, then the second row, then the third, etc. And each row would contain 1-bar split into 16 sections.

So, the above would sound like…

:notes: Out__of_____time. Out_of_______time. Out____of_time. :notes:

There could be 4 rows, and in addition 16 “snapshots” of rows that you could control via CV. (Essentially, the snapshots would act like a built-in Stoermelder 8FACE .

As for word selection, we could use existing research (such as http://cs.brown.edu/courses/cs100/students/project11/) to choose 128 words or so to make available.

I could synthesize the words - Or maybe we could let people create “packs” of recorded words? A pack would contain a folder of samples and a file that lists what words are available and which sample is associated with which word, and how long the magnet should be for each word. Very interesting!

I created another mock up to show that there would be a “pool” of words to choose from:

6 Likes

Hi, Bret,

that looks very interesting and to me quite understandable.

:+1: cheers

1 Like

that’s definitely sounds cool! Wanna test it already) it’s something new in this area)

1 Like

I love this idea, with visuals like that I could introduce it into maths and english lessons and we’d have way more fun.

I know this is not easy to work out, but grammar features leading to some weird sounds would be nice - but I could use that with some distortion module and ask them how deep we need to destroy the sentence coming out depending on the amount of mistakes they spot :smiley: (or if they love distortion reward them with higher distortion freedom by level of grammatical correctness).

Lucky me we only have to build very small sentences as the vocabulary capacities of my non native english speaking learning impaired children are low.

1 Like