@amir @nick Hmm, I don't hear the sizzle so much in my version. I also wouldn't know how to make it better now. Bringing the sibilance component back would be too risky for me, because you didn't like that either. On the other hand, I might try to do that and decrease the overall s gain instead afterwards. But I don't know.
Sep 5, 2026, 8:46 AM@amir @nick ok, here is a test version I put on dropbox. If this still sounds ok while fixing the old man issue, I will publish this as v21.3
https://www.dropbox.com/scl/fi/i7ln8r4s3b7kqo604srif/Eloquence-v21.2-test-s-balance2.nvda-addon?rlkey=z44o3pvgsc0urtqaar23vzkje&dl=1
Sep 5, 2026, 9:52 AMSep 5, 2026, 10:56 AM@amir @nick yes, I got that. That's why I didn't release it. And you mean every s sound? I'll stick to v21.2 for now, as I don't have an idea how to improve it. I checked it multiple times against a recording of the external 22 kHZ output, and the differences seem to be very marginal now, if at all there. I will improve it further when I have an idea, but for now, I've run out of ideas.
Sep 5, 2026, 11:24 AM
Sep 5, 2026, 11:26 AM@amir See, I sort of disagree with @HOzosch here. At least in my Rust host, You're getting the EQ for free, performance-wise. The gap between when you switch from 11 KHZ to 16 KHZ is it literally live replacing files.
Sep 5, 2026, 11:42 AM@nick Here the problem lies with the 16 kHZ and the TTS switch delay I guess.
Sep 5, 2026, 11:44 AM@amir Yeah, and that's why I'm going to end up having to double the addon's size, as I'm going to have to store two copies of every language so it's true switching, not live patching.
Sep 5, 2026, 11:48 AM@nick Quite understandable and logical.