But anyway, to actually answer your question, you don’t necessarily need FM to do rudimentary speech synthesis. The Monomachine manual actually rather nicely explains how to do a basic version. You need to consider vowels and consonants: vowels are long and pitched sounds, and the individual vowel sounds are determined by about two formant filters (you can try bandpass or near equivalent resonant lowpass filters with specific resonance frequencies that match the resonant frequencies of the major human vocal tract cavities. For compounded vowels, you have to slide from one vowel to another. For consonants you can use noise or other unpitched sounds.
The difficulty for realistic speech is that people adapt their sounds depending on what the preceding and subsequent sounds are. It’s very difficult to emulate that with a sequenced musical instrument, which is why most speech synths sound so stilted. I don’t have an AF, but I know it would be extremely difficult to do. Why not start with some sustained vowel sounds?
I made a pretty convincing opera singer on the A4, I’ll try to dig up the sound and post it here…
PeterHanes is right on, the 2 resonant peaks close to each other are key to vowel sounds. The 2 lowpass filters of the A4 with each a resonant peak at slightly different freqeuncy should do the trick. Add a little vibrato and there is the opera singer 
couldnt help myself but need to post this here
Erm, I think I was wrong. I think that it’s jonah who is writing the speech synth app for AF.[/quote]
i messed with it a bit, it really wasn’t worth the time imo. i find it’s easy enough to get vocal like sounds and kinda more interesting and faster to use in music to do it by hand because i don’t want intelligible voice, just vocal sounds.
beyond that i’d just buy this on the cheap
i guess yamaha has software you can grab to make it say whatever.
i wonder with the fpga if the insides of the a4 are reconfigurable enough that maybe someday elektron will reconfigure the a4 to work as a vocoder.
They made a sex toy talk. That cracks me up.blahshahahhhhhhhh
Hi dear elektronauts
i just wanted to know if its possible to get a voice like sound out of the A4, as heard often in Psytrance, Progressive Goa and the likes…
for example: https://www.youtube.com/watch?v=FHPZDSisKaM
around 4:00 minutes in.
and if so, can u tell me how thanks in advance
I did some work on making nice vocal formants on the A4 a while back. i put together a little cheat sheet for the frequencies I needed and then went to town making a series of presets. Heres what i was working off of:
https://dl.dropboxusercontent.com/u/1303726/formantnotes.pdf
My approach was to create each one of those vowel formants using two chained together voices of the A4. I set the first track to have 100 volume in the amp section but 0 track level then on the next track I set osc 2 to neighbor and turned Bosc 1 off. On both tracks I opened filter 1 100% and set filter 2 to peak mode. the cutoff settings in the A4’s filters actually correspond to midi note numbers so you can dial in specific frequencies. The chart on page 2 in the PDF above shows you which note numbers correspond to which frequencies. For each vowel I set track 1’s f2 cutoff to the lower frequency and did the same on track 2 for the high frequency. I saved both as a preset with the name of the vowel and the track number. There are some tables out there with a 3rd formant frequency for each vowel so I added a 3rd track to all the presets at some point. When I finished I added all of them to the sound pool. Now it’s easy to create a melody on the sequencer, copy the trig pattern to multiple tracks, and use sound locks to get to vocal sound. I also set up the performance controls to offset the f2 cutoff of both tracks simultaneously making formant shifts if they’re adjusted by the same amount or making a formant stretching sound if they’re different. Here’s a dumb little pattern i threw together to show you what it sounds like.
That sounds way cool man; thanks for sharing
Helloliamdaly, cool sounds!
I actually really like making voice-like sounds on the A4. I usually go for oscillator sync and a peak filter, with some carefully modulated noise. I could post a more detailed explanation tomorrow if anyone is interested.
Anyway, I made a really silly cover of the silly German version of the opening theme for the classic Sailor Moon anime - using only the Analog Four, of course. 
https://soundcloud.com/i-i-i-o/deutsche-moonlight-densetsu
So, the voice there is my patch.
EDIT: OH, okay - not THAT voice like. I misunderstood.
For that I would go for some distortion and again, the peak filter. I think there is a TB-303-ish preset on the A4 which would sound very similar if you tweak the attack of the filter envelope.
Here’s a sysex with the sounds. They’re named with the track they’re supposed to go on. Again, the easiest way to use them is to add them to the sound pool, write a melody on track 1, copy the pattern to tracks 2 and 3, then do your sound locks on each trig across all 3 tracks to get the formants going.
Side note - there may exist an easier way to do this in unison mode with play track sounds enabled but I haven’t gotten that to work well for me.
If you want to alter the melody or add vibrato you only need to change track 1. You can adjust volume and amp envelope settings on track 3. For a formant shift performance control set a performance macro to apply a negative amount (around -25 works for me) to f2 cutoff on all 3 tracks. For formant stretching do the same but apply different amounts for each track.
Since formants have more than one amplified peak in the spectrum of harmonics, I use synths with two filters and with two different resonant settings.
With the A4 this can be done with one synth voice and the two filters. If the cutoff frequencies and the resonances are set right, a “vowel”-like sound will be generated. If the cutoff frequencies and/or the resonances are modulated, formant-like transitions can be generated.
Listening to the track of the OP I would guess, it has been done with a 24dB LP filter that has a beefy character by itself and it’s cutoff was slowly modulated and with moderate (50-60%) resonance. I think a Moog filter (Voyager) would sound like this, if feeded with a 32’ VCO, without much of an effort. Just checked this out and it was surprisingly close to the example.
But on the A4 I would try the multimode filter, used as LP, for the resonance, because it is much more prominent, than the 24db LP … the 24dB LP could be useful to get rid of the higher harmoncs, because the mulitmode is 12 dB only.
But it’s only a guess … and there are so many ways … 
If I still had an A4, I would try routing one voice into the filter of a neighbor track, use modulated notch filters on both tracks, and FM the filters with the new tunable LFOs. Aren’t formant filter usually a series of notch or bandpass filters?
hey folks,
thank you for the tipps but yeah i think im too stupid to get the sound im searching for 
i played around with pitch mod / freq lfo’s the filters, separated and combined, using 1,2,3 or all 4 tracks in unison, reset, rotate. I changed envelopes and osc but i can’t find the correct recipe… (for the last 2 weeks trying)
im feeling kinda dumb T-T so i call out in despair for a idiot proof solution…
Thanks, that is helpful information 
Wait … it should be easy, really … if you are searching this deep growling voice. I can’t say that this has been done with an A4, because I would make my bet on a Moog (Voyager or one of the Phatties) … this smudgyness sounds very familiar to me. I think, one track and one or two filters could do the job on an A4. Maybe tonight I have some time to try it out myself, because I like this sound too!
This would be my approach…
- get a bass note, C3 or C2
- use a LP filter with decent amount of resonance, but not going into selfresonance, you don’t want to get too much of squelching right at the beginning
- set the influence of the envelopes to zero to work with sustained sounds
- now play the bass note and listen, while you’r moving the cutoff-frequency - manually - and slowly
- if you get some singsong while moving the filter cutoff, that seems similar to your sound, you are beginning to hitting home
- consider that the LP (first filter) has less characterful shaping of the sound then the second Multimode filter. You can use both and emphasize the effects of resonance.
Now, if the gritt is not smudgy enough, you can apply FM-modulation to the filter-cutoff with very low - but suited - frequencies. Best would be to use subharmonics of the original note and get this FM following the keys.
I hope that can give you some direction. Your desribed approach is almost there …
EDIT:
Well, have tried it myself now … honestly … seams not that easy as I thought. My approach seams to go in the right direction - but - I got something that I like and I will use it, but it’s not the same sound. Maybe the A4/AK filters just sound somehow differently … 
Could it be that this sound has been made with a vowel-filter - or - at the end - could it be a sample played back at reduced speed? A belching male voice for example?
Nice little riddle - this is
seriously coo thread!
Big thanks to everyone who shared their patches and notes! So cool to know how the Cutoff Midi values correspond to actual frequencies!
Please do! I’m also looking to do some psytrance style formant filter-esque voice sounds on the A4. This is a good start
Here is a good example of what I’d be looking for…
Edit: That one that comes in at 3:41 especially!
Interested in this?
Native-Instruments/Reaktor/Razor/Leads/Percuvox, played at C2
It’s not the same follow-up of vowels, but from the timbre it seems quite similar. The sound is made by a narrow PW and the VOWEL-filter. Even if it’s not the same, Razor has some very intersting gritty and wet sounds that would gladly fit in the audio example of the OP.