English


Profile Picture

Voice.AI voice changer

Posted By General Picture Animation 3 Years Ago
You don't have permission to rate!
1
2
3
4
5
6
7

Voice.AI voice changer

Author
Message
Sunglass
Sunglass
Posted 2 Years Ago
View Quick Profile
Distinguished Member

Distinguished Member (1.7K reputation)Distinguished Member (1.7K reputation)Distinguished Member (1.7K reputation)Distinguished Member (1.7K reputation)Distinguished Member (1.7K reputation)Distinguished Member (1.7K reputation)Distinguished Member (1.7K reputation)Distinguished Member (1.7K reputation)Distinguished Member (1.7K reputation)

Group: Forum Members
Last Active: Last Year
Posts: 258, Visits: 1.3K
This guy knows about audio


Pollux
Pollux
Posted 2 Years Ago
View Quick Profile
Distinguished Member

Distinguished Member (2.6K reputation)Distinguished Member (2.6K reputation)Distinguished Member (2.6K reputation)Distinguished Member (2.6K reputation)Distinguished Member (2.6K reputation)Distinguished Member (2.6K reputation)Distinguished Member (2.6K reputation)Distinguished Member (2.6K reputation)Distinguished Member (2.6K reputation)

Group: Forum Members
Last Active: Last Year
Posts: 838, Visits: 1.8K
Dommage que les francophones ne s'intéressent pas à l'IA pour la voix


Si Charly il y a pas mal de français qui s'y intéressent malheureusement ils restent dans ''l'ombre''...!

charly Rama
charly Rama
Posted 2 Years Ago
View Quick Profile
Distinguished Member

Distinguished Member (8.1K reputation)Distinguished Member (8.1K reputation)Distinguished Member (8.1K reputation)Distinguished Member (8.1K reputation)Distinguished Member (8.1K reputation)Distinguished Member (8.1K reputation)Distinguished Member (8.1K reputation)Distinguished Member (8.1K reputation)Distinguished Member (8.1K reputation)

Group: Forum Members
Last Active: Last Year
Posts: 1.9K, Visits: 17.3K
Yes, it's really interresting, but I would like to see french language support. I tried, it works but it keeps english accent. Dommage que les francophones ne s'intéressent pas à l'IA pour la voix


     


mtakerkart
mtakerkart
Posted 2 Years Ago
View Quick Profile
Distinguished Member

Distinguished Member (16.8K reputation)Distinguished Member (16.8K reputation)Distinguished Member (16.8K reputation)Distinguished Member (16.8K reputation)Distinguished Member (16.8K reputation)Distinguished Member (16.8K reputation)Distinguished Member (16.8K reputation)Distinguished Member (16.8K reputation)Distinguished Member (16.8K reputation)

Group: Forum Members
Last Active: Last Month
Posts: 3.2K, Visits: 29.2K
It start to be very interesting


Turjan
Turjan
Posted 2 Years Ago
View Quick Profile
Veteran Member

Veteran Member (843 reputation)Veteran Member (843 reputation)Veteran Member (843 reputation)Veteran Member (843 reputation)Veteran Member (843 reputation)Veteran Member (843 reputation)Veteran Member (843 reputation)Veteran Member (843 reputation)Veteran Member (843 reputation)

Group: Forum Members
Last Active: Last Year
Posts: 47, Visits: 365
Ooh, looks like I missed this thread!

It's quite relevant to the videos I make because I'm primarily a "one man band" when it comes to the actors in my stories. Sometimes I'll co-opt a friend into recording some dialogue for me, but mostly it's just me voicing the characters.

The exception is the A.I. crew on my main character's spacecraft - they're voiced by Ivona text-to-speech software. For all the other characters voiced by me I use MorphVOX Pro, and in a lot of cases all it takes is slight shift in pitch and/or timbre to create a different enough voice, especially when it's tied in with speaking in a different accent.

Both tools have their pros and cons, but weirdly the biggest problems with each stem from completely different causes...

With a tool like MorphVOX Pro, I find that a lot of the problems folks I've talked to have comes down to what their expectations are of what the software can do, rather than what it's actually capable of. MorphVOX (and similar tools) change pitch and formants, but the voice is still yours, meaning that the inflections, delivery and pacing of your way of talking remain unchanged. In order to create a voice for a different character you have to change your speech pattern, i.e. you have to act the character yourself. This is what makes a tool like MorphVOX flexible.

The text-to-speech (TTS) Ivona voices have the opposite problem : their chartacter is fixed. You feed them words and they say them how they say them, no flexibility involved. So if I want my Ivona voiced characters to say things a certain way, I have to change the script, tweaking words until the software delivers a line in the style I'm looking for. My A.I. characters always sound artificial, but that's fine because in my stories, that's exactly what they are - Artificial Intelligences.

There is a third class of software now of course, that's essentially a hybrid of the two, and this is the category that Voice.AI and software like VoiceMod fall into. This type of tool uses samples of a person's voice to essentially train an Ivona-like filter, only more flexible. Such filters can handle change such as using a different accent, while still retaining the essence of the original person's voice.
In theory anyway...
In practice however the tech is still in its youth, and the versions available to joe public at an affordable price tend to produce voices that have a lot of quality issues and still often suffer from that hint of roboticism we associate with TTS voices.

I'm curious to investigate the possibilities of these new voice changers, but until the technology improves a bit, I'll be staying with my existing tools thanks!

By way of illustration, here's a link to a video I did a few years - and a few iterations of iClone - ago (the characters look so much better now with SkinGen shaders!). In this video, all the voices are me, with the exception of the two female A.I. characters Cass & Andra. All my audio tidying up was done in Audacity.



Also, here's a little thing I knocked up specifically for this post. I had been intending to just upload an .mp3 to illustrate the VoiceMod filters, but as the forum doesn't seem to like .mp3s, I thought "Eh, what the heck, I'll turn the audio file into an AccuLips animation in iClone8 instead. I've been wanting to try out the new expression wrinkles anyway!"
So, this hastily assembled video shows a couple of the A.I. voices in VoiceMod and points out some of their good points - and their limitations...



bob_441345
bob_441345
Posted 2 Years Ago
View Quick Profile
New Member

New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)

Group: Forum Members
Last Active: 2 Years Ago
Posts: 2, Visits: 2
No, but I did search for forums and this seemed to be the only one where it was being discussed that I could find, so I figured someone would be familiar with it and might be able to help. Thanks you.
animagic
animagic
Posted 2 Years Ago
View Quick Profile
Distinguished Member

Distinguished Member (33.4K reputation)Distinguished Member (33.4K reputation)Distinguished Member (33.4K reputation)Distinguished Member (33.4K reputation)Distinguished Member (33.4K reputation)Distinguished Member (33.4K reputation)Distinguished Member (33.4K reputation)Distinguished Member (33.4K reputation)Distinguished Member (33.4K reputation)

Group: Forum Members
Last Active: 2 Weeks Ago
Posts: 15.8K, Visits: 31.3K
Bob, voice.ai is not a Reallusion product. Did you try to get help from voice.ai directly?



bob_441345
bob_441345
Posted 2 Years Ago
View Quick Profile
New Member

New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)New Member (2 reputation)

Group: Forum Members
Last Active: 2 Years Ago
Posts: 2, Visits: 2
I am a brand new user of voice.ai and wondered if anyone can walk me through how to use the software for my application. I want to create my own voices, and then load those into the system so I can access them. I've already uploaded one voice and got the email saying it was complete. I included over 2 hours of pre-recorded speech. It does show up in my Uploaded Voices section, but when I want to use the microphone to speak in that voice either live or recorded, I can't find a way to select the voice I created. I don't see any way to search for voices either. Are there any really detailed tutorials on setting up voices and using them in this way. I'm really only interested in using the recorded voice feature, not the live feature.  Thank you! I hope someone can help me. I guess I'm too new here to start my own topic yet.
AutoDidact
AutoDidact
Posted 3 Years Ago
View Quick Profile
Distinguished Member

Distinguished Member (6.1K reputation)Distinguished Member (6.1K reputation)Distinguished Member (6.1K reputation)Distinguished Member (6.1K reputation)Distinguished Member (6.1K reputation)Distinguished Member (6.1K reputation)Distinguished Member (6.1K reputation)Distinguished Member (6.1K reputation)Distinguished Member (6.1K reputation)

Group: Forum Members
Last Active: 2 Weeks Ago
Posts: 2.1K, Visits: 13.6K


cepheistudios (12/25/2022)
Yes. That would give you the 60 minutes. But, doing that much voicework in a month (unless you're allocating a lot of time to it) is extraordinary - which is something I touched on before (as Mike Kelley). Let me explain why: I did my own voicework for a 2 minute long Youtube piece. I wrote the script and you might be thinking it should be easy. It was ANYTHING but quick. It took thirty minutes. I would mis-pronounce words. I'd mess up the cadence. My mouth would get too dry in the middle of the take. The mic wasn't adjusted properly so it was too quiet. Lots of things happen even when you're in control of every element. And that was using my own voice without any alterations at all. Now imagine how much time it adds to change voices, pitch, tone, etc. Imagine you do a take and you think the AI voice great but, maybe it isn't as great as you thought after you hear a take. You'd be really hard pressed to use all 60 minutes in a month. It's definitely possible but read this: https://ryderecordingstudio.com.au/tech-articles/how-long-does-it-take-to-record-a-voiceover/


I agree that even with a really good voice changer that doing a Feature length film ( 60+ minutes), with multiple characters, all by yourself ,is extremely labor intensive.

in fact I do not recommend it and will never try such a task again after my 93 minute “Galactus” Rising” film that took six years to complete.

for that film I used the horrible morph vox pro.and studied various foreign accents& dialects.

At some point it becomes too much to keep track of.

https://youtu.be/Xja0DNS1rRc





RAG DOLL COLLISION ANIMATIONS FOR ICLONE 8 & 7
---------------------------------------------------------------------------------------------------------------------
Ghost Origins
My latest Feature length film created with Iclone.
76% of original size (was 661x19) - Click to enlargehttps://forum.reallusion.com/uploads/images/adf9b210-df59-4cb6-aa1b-9de5.jpg
My Sci- Fi Graphic Novel on Amazon: https://a.co/d/9k3cwoY


Kelleytoons
Kelleytoons
Posted 3 Years Ago
View Quick Profile
Distinguished Member

Distinguished Member (37.8K reputation)Distinguished Member (37.8K reputation)Distinguished Member (37.8K reputation)Distinguished Member (37.8K reputation)Distinguished Member (37.8K reputation)Distinguished Member (37.8K reputation)Distinguished Member (37.8K reputation)Distinguished Member (37.8K reputation)Distinguished Member (37.8K reputation)

Group: Forum Members
Last Active: Last Year
Posts: 9.2K, Visits: 22.1K
Well, since we're sharing...

I used the NyVox beta (or perhaps it's even the alpha - whatever it is, you can use it via Patreon for $10 a month).  It has 200 voices (now) and some of them sound very good... for short sentences spoken slowly.  That convinced me to try it further, butt for an animation project I was trying to finish for Christmas yesterday.

It destroyed my time table.  Granted, I was doing a LOT of processing with it (four voices, each around a minute of dialog) but the problem was not speed (it operates in real time, more or less, as the avowed purpose is for folks to use in gaming and online events) but quality.  As the previous poster said, this software also had problem with words, clipping some of them or just mispronouncing so you had no idea what the real word was (from an MP3 that was very well processed and done - the source was NOT the problem).  After a few hours I realized it just wasn't even close to something I'd ever use in an animation (so I'm going to end my support of it - perhaps they will get it right, but I now have zero confidence in it).

I KNOW there is software that works right - Corridor Crew used some - but it's VERY $$$, and way out of the range of most of us.  Plus if you want to create your own voices (which I think is the main purpose of it - I don't remember seeing voices offered, unless they were ALSO $$$) it takes a LOT of training time.  AND it also takes a lot of processing power (again, to do it right - this won't be happening in real time.  Which is okay for us animators).

Based on what I've seen and read I'm not sure we're even close to having something affordable for serious animation work for us hobbyists.  



Alienware Aurora R16, Win 11, i9-149000KF, 3.20GHz CPU, 64GB RAM, RTX 4090 (24GB), Samsung 870 Pro 8TB, Gen3 MVNe M-2 SSD, 4TBx2, 39" Alienware Widescreen Monitor
Mike "ex-genius" Kelley

1
2
3
4
5
6
7



Reading This Topic

0 active, 0 guests, 0 members, 0 anonymous.
No members currently viewing this topic!