Well, if anyone ever gets me a copy of SVCTL32.dll / SVCTL64.dll without the timer restrictions, would seriously pay like a hundred bucks for it. But it's never gonna happen. Sigh. That DLL is forever time-locked to something Image Line do in FL Studio to give it proper licensing. So no newer softVoice, no 64-bit SoftVoice build until then. The DLL files at eurpod.com/svfl.zip for those interested. SoftVoice as a company is quite defunct, their website is up yes but as far as a company with filings and public record they have not been around since the mid-2000s. So, not a chance of reaching out and asking for my own licensed copy either. This is all making me very sad, now I've lost all motivation to code.
I tried to see if it was returning the last chunks of text in the callback, but nope. SVAudioCallback is either a query function (not a “register callback” function), orSVCTL only enables callbacks after some other registration step (likely via SVRegister / SVCommand), or callback is only for a different output mode (like “send audio to file” or “external device”), and isn’t used for TTS the way we’re invoking it. That's why my hunch here lies with the demo restriction and a timer that is set for each utterance, that when stopped throws away the last chunks of spoken feedback.
I tried to see if it was returning the last chunks of text in the callback, but nope. SVAudioCallback is either a query function (not a “register callback” function), orSVCTL only enables callbacks after some other registration step (likely via SVRegister / SVCommand), or callback is only for a different output mode (like “send audio to file” or “external device”), and isn’t used for TTS the way we’re invoking it. That's why my hunch here lies with the demo restriction and a timer that is set for each utterance, that when stopped throws away the last chunks of spoken feedback.
This entry was edited (5 days ago)
Timothy Wynn
in reply to Tamas G • • •Jayson Smith
in reply to Tamas G • • •Tamas G
in reply to Jayson Smith • • •FL (or the Speech Synth plugin/module) calls ZXSpeakInit in speaker.dll.
Then, speaker.dll loads/uses svctl (“control” layer), and svctl loads/uses sveng (the actual voice engine). (This is implied by the dependency on SVCTL32.dll and the “3 file” requirement.)
speaker.dll then asks SoftVoice to synthesize the text, likely into a PCM buffer.
Finally, speaker.dll (maybe via MSACM32.dll) converts/normalizes the audio into a format FL expects.
So in theory inspecting details within speaker.dll can give all the clues too, but reading disasmed dll is so not een in my pervue, lining up function names to areas of the DLL, loading it in debuggers to check what it dos in memory, ETC...
True about DecTalk though. It's why I was surprised when Blazie used it in BTSpeak, though perhaps there's some other licensing / conversations we don't know of there so I won't assume anything. Still, it's either that or clear they saw low-risk at distributing it more commercially within BTSpeak units. Very interesting. And yet, Apple themselves probably will never touch it, despite it being so in the open in sources even if unofficially, they just know they can't touch it with the historical significance and potential licensing rights for them.
Jayson Smith
in reply to Tamas G • • •Tamas G
in reply to Jayson Smith • • •Jayson Smith
in reply to Tamas G • • •Kara Goldfinch
in reply to Tamas G • • •