[Update info] Environmental sounds, de-cringed voicing system preview, redesigned say screen, Soumya, big fantasy for Natalie/Rebecca style content fans and more
Haven't slept for 36 hours since I managed to pull this off (just applying it has been super fun), writing this before hitting the bed. Upcoming update will be pretty fat, as it stands now an increase of 600 MB featuring stuff for Soumy fans, fans of Natalie / Rebecca style content and the other stuff that's been voted on.
This post details something that will be showcased for the first time ever in a few prototype scenes. The return of sound, environmental and voices
Clacking of heels, bustling streets, handcuff sounds, ropes, car doors opening, car doors closing, car engine starting, elevator doors opening, closing, beeping, whirring, elevator music etc. 6 separately controlled and layered channels + 7th voices. They help hide any imperfections in background noise of voices, but also help ground scenes. It adds depth to it.
But, most importantly:
VOICES. ACTUALLY GOOD VOICES.
Now, OGs might be cringing and rightly so.
Initially, we were dependent on super expensive and super garbage ElevenLabs stuff which had zero proper pacing, zero emotional expressiveness and most importantly zero context, making it cringe.
There's been large strides in open sourced alternatives in the last year and I've been working on creating a specialized "platform" on top of one of these engines to facilitate Bindr's dialogue system.
I've expanded it with preparsing and annotations which link it directly to dialogue in the game by mapping the character to the voice file (each generated through prompting and then refined and mixed until it works every time). Also added support for the system to have more context about the scene and what is going to be said to more effectively generate reliable lines (combined from many full attempts).
I then distribute the work of a whole dialogue to be generated on two 24 GB GPUs (required for this) and begin to generate large outputs of 2-3 minutes of dialogue at a time at a speed and quality which puts ElevenLabs v3 to absolute shame. Then proceed to recombine them, the good parts, throw away the bad. It's still highly stochastic.
I let it run wild and then find each line in sequence to stitch into final output. Yes, this takes quite a bit of effort, but sometimes it's really worth it. OGs might also remember the voice of the first girl of Bindr in 2023, Rebecca, that you can hear in its proper contextualized glory in the audio attached.
[Rebecca] Interrupt me one more time and you will be soundly flogged.
[Secret] Yes, mistress.
[Rebecca] Good. Now, last time you came here, you were quick to escape. What's changed?
[Secret] Well, at first, I was caught off guard. But the more I thought about it... I felt aroused?
[Rebecca] Oh? Do tell. A good submissive origin story might save you.
[Secret] I... I couldn't stop thinking about your whipping post, fantasizing about being restrained to it, anticipating your lashes.
[Rebecca] A pain slut and a rope bunny. Hmm, I do relish the sight of you in that predicament.
[Rebecca] I do not usually train a sub with a whip on day one, but your behavior warrants it. Are you willing, slave?
[Secret] But I don't know if I'll actually be able to take the pain...
[Rebecca] Once you're restrained, it won't matter. The cuffs will keep you in place.
[Rebecca] Now, choose two pairs of handcuffs from the wall. Put them on your wrists and assume the position.
[Secret] Shouldn't I take my clothes off?
[Rebecca] We have scissors, darling. Well, I do. Soon, you will have nothing but shame and humiliation.
[Rebecca] Understood?
[Secret] Yes, mistress.
[Rebecca] Excellent. Now walk! Every minute you delay is a lash added to your sentence.
And yes, normal conversations work, funny quips work, laughing works, sighing works, flirty conversations are not cringe etc. Also sexual stuff works, heavy breathing, moaning, humming singing. Just takes work, but it's finally possible.
Redesigned say system
By moving who says it to the corner, it takes up less space from the image. I will still probably leave a toggle to go back to old style. But I personally greatly prefer it. Feels more movie-like.
Soumya and everyone else voted on is coming
I made a promise with the voices when I restarted the project, and I like to keep them. Same with voted on stuff. I'll also have the voting on the next series of girls for September after I put out the full update.
I don't want to spoil too much so avoiding spilling the beans.