Training someone to change the perceived gender of their own voice is a similar process to training them to speak with a new accent. I recently prototyped a gender-affirming voice therapy app for transgender women: https://own-voice-dev.netlify.app/
So now I’m curious if our two applications use the same approach under the hood! I have a detailed tech stack rundown on this page: https://own-voice-dev.netlify.app/#/info but in brief I use OpenVoice v2 to convert someone’s voice toward a differently-gendered reference voice, but leave it at only like 30%-60% transformed so it still sounds like them and not the reference. This transformed voice then becomes a practice reference for drilling.
I know it’s bad form to discuss that, but this is a completely normative comment, with no opinion given so far from being a flame. And it adds to the technical discussion!
Probably should make sure you're thinking about your actual UX when your third party voice cloning service cuts you off. Not a great sign when I can't even go through the demo without hitting a back end error from a third party.
I got ```
Voice cloning failed (403): {"detail":{"type":"authorization_error","code":"forbidden","message":"You have reached your monthly limit of voice add/edit operations (95). Please consider upgrading your subscription to increase your
```
Definitely something that you should have made sure doesn't leak into the front end UI.
Voice cloning failed (403): {"detail":{"type":"authorization_error","code":"forbidden","message":"You have reached your monthly limit of voice add/edit operations (95). Please consider upgrading your subscription to increase your limit.","status":"voice_add_edit_limit_reached","request_id":"zzz"}}
The only accent I treat differently is the unsolicited spam call.
Face to face, I work with Indians, Ukrainians, Vietnamese, Mexicans, Brazilians, Russians, Guatemalans, Cubans, Italians and I love to hear the twang each person has.
Curious what you mean by treated differently or why you needed to put so much effort into hiding your accent? Are you sure it wasn't just general English? Because of course you may get treated differently if you can't speak English...but I think the accent is not really a problem.
Cool. Thank you for testing it. It can have multiple purposes. For people that want to do video presentations for instance and have a more native accent. Or yes, for a podcast. What do you think I should improve?
This is good, although it left a fairly antipodean tone with a nerdy style when I selected British English. The source was a fairly heavy french accent, with the outcome being quite amusing.
The token limits seem to have kicked in now (my first go was a few hours back and this is just my second attempt)
Training someone to change the perceived gender of their own voice is a similar process to training them to speak with a new accent. I recently prototyped a gender-affirming voice therapy app for transgender women: https://own-voice-dev.netlify.app/
So now I’m curious if our two applications use the same approach under the hood! I have a detailed tech stack rundown on this page: https://own-voice-dev.netlify.app/#/info but in brief I use OpenVoice v2 to convert someone’s voice toward a differently-gendered reference voice, but leave it at only like 30%-60% transformed so it still sounds like them and not the reference. This transformed voice then becomes a practice reference for drilling.
I am mostly using elevanlabs for this
I’m surprised this comment was downvoted.
I know it’s bad form to discuss that, but this is a completely normative comment, with no opinion given so far from being a flame. And it adds to the technical discussion!
Probably should make sure you're thinking about your actual UX when your third party voice cloning service cuts you off. Not a great sign when I can't even go through the demo without hitting a back end error from a third party.
I got ``` Voice cloning failed (403): {"detail":{"type":"authorization_error","code":"forbidden","message":"You have reached your monthly limit of voice add/edit operations (95). Please consider upgrading your subscription to increase your ```
Definitely something that you should have made sure doesn't leak into the front end UI.
shit. it should be fixed now. Can you try again? Will fix the leak too
Voice cloning failed (403): {"detail":{"type":"authorization_error","code":"forbidden","message":"You have reached your monthly limit of voice add/edit operations (95). Please consider upgrading your subscription to increase your limit.","status":"voice_add_edit_limit_reached","request_id":"zzz"}}
I think we burned up your tokens
haha. was not expecting this amount of traction. Should be working now
Interesting, what's the purpose? Maybe, host a podcast without revealing your real voice?
I tried a few, I think the British American was maybe the best one. I didn't really notice a huge difference on Italian or Indian accent.
British - American https://pastewaves.com/player/25cacf64-461a-4eef-9b02-9bc0da...
A little shaky, but no doubt hints of a British Accent.
I went to a lot of work to adopt a “neutral” American accent, with some success. Perfectly adequate for casual conversations.
Why? You run across many people with accents every day, right? Well, it’s a problem; people definitely treat you differently.
This issue isn’t specific to the USA btw, it’s just I’ve been working in the USA for a long time.
The only accent I treat differently is the unsolicited spam call.
Face to face, I work with Indians, Ukrainians, Vietnamese, Mexicans, Brazilians, Russians, Guatemalans, Cubans, Italians and I love to hear the twang each person has.
Curious what you mean by treated differently or why you needed to put so much effort into hiding your accent? Are you sure it wasn't just general English? Because of course you may get treated differently if you can't speak English...but I think the accent is not really a problem.
Cool. Thank you for testing it. It can have multiple purposes. For people that want to do video presentations for instance and have a more native accent. Or yes, for a podcast. What do you think I should improve?
This is good, although it left a fairly antipodean tone with a nerdy style when I selected British English. The source was a fairly heavy french accent, with the outcome being quite amusing.
The token limits seem to have kicked in now (my first go was a few hours back and this is just my second attempt)
It looks cool, but how do I know this is not a front to capture people’s voice fingerprints for other sinister usage?
hmmm... What type of sinister usage?
Well not giving you ideas but voice signatures or prints are used in the wild, in wars or even civil cases, even in the US https://www.technologyreview.com/2019/01/31/137633/us-prison...
interesting. giving me a lot to think about
can it do brummie?
what is brummie?