Wrong Gender Voice

[SyamsQ]

Here I have two Indonesian female voices.

However, when I use these voices for dubbing, the resulting voices are male.

[daveshenal]

In your second screenshot, look at the "Designed Voices" panel: Wanita M" shows [object Object] where the identity tags should be(compare it to "Voice 07:17 PM - DESIGN" right above it, which correctly shows female, young adult, high pitch, british accent).

That [object Object] means Wanita M's gender/identity data wasn't saved as a normal string. It's a stray JavaScript object that isn't being read correctly. When that corrupted profile gets sent to the TTS engine at synthesis time, the engine likely can't parse the gender field and falls back to a default voice, which seems to be male.

As a workaround for now: delete the "Wanita M" profile and recreate it through Voice Design (set gender/age/pitch fresh, then "Save design as profile") rather than however it was originally created (imported, cloned, or hand edited). A freshly saved profile should store identity as a normal string and avoid this.

For the maintainer: it'd help to know how "Wanita M" was created (cloned from a sample? Imported as a .ovsvoice bundle? Edited via API?) since that'll narrow down where the object is leaking in instead of being flattened to a string before save. A diagnostic bundle (Settings -> Diagnostics) attached to this issue would also help confirm whether it's a frontend display bug only or whether the bad data is actually what's reaching the TTS engine.

[debpalash]

@daveshenal called it exactly right — the [object Object] in the identity tags was the tell. 🎯

What happened: a bug in an older build saved the designed voice's gender/identity data as a raw object instead of the tag string. An early cleanup pass then blanked the corrupted value — which silently threw away your design, so "Wanita M" rendered with the engine's default (male-sounding) voice instead of your female design. Exactly what you saw.

This is fixed since v0.3.7 (PR #600, which cites this exact Indonesian-female-voice case): the app now rebuilds the identity from your design's saved category picks at synthesis time — your existing voices heal automatically, no need to re-create them — and corrupted identity data can no longer be saved in the first place.

If you're on the latest release (v0.3.10) and a designed voice still renders the wrong gender, that would be new information — please open an issue with the voice's design settings and I'll dig in. Marking this answered; thanks for the clear screenshots, and thanks @daveshenal for the sharp diagnosis.