logo Sign In

Jar Jar Bricks

User Group
Members
Join date
15-Jun-2019
Last activity
1-Jul-2025
Posts
2,933

Post History

Post
#1523110
Topic
The Rise of Skywalker: Ascendant (Released)
Time

Bad news everyone: a lot of the Poe lines have a comm filter placed over them, and he’s using a shouty voice in like 90% of them, too. I used only the ones that didn’t have the filter, and the results are pretty subar. Definitely not passable for a monologue. If anybody knows where I can get more content of Oscar Isaac, please let me know. It would need to have no background noise or other people speaking.

Post
#1523100
Topic
The Rise of Skywalker: Ascendant (Released)
Time

Here are 3 more “become” line options. I shortened it to just “we’ve become a dyad in the Force”, so you could sandwich it between What “Palpatine doesn’t know is…” and “Rey”. Or you can simply edit out the “dyad in the Force” part with Audacity.

Option 1 (kinda crap): https://drive.google.com/file/d/1T4dov0e3mAEU-FqizUDpGZHA2EEeHwap/view?usp=sharing
Option 2 (decent): https://drive.google.com/file/d/1ossG8GEGG-oIlOkxcuHQj9Lm4Lxn7BK5/view?usp=sharing
Option 3 (This is my preference because you can hear Adam’s voice very clearly under the mask): https://drive.google.com/file/d/1K7mkjHw2Avt7GpQX6RxFe8I_HjVFGrR4/view?usp=sharing

Post
#1523077
Topic
The Rise of Skywalker: Ascendant (Released)
Time

I’ve heard people trying to replicate Alec Guinness on YouTube. It never sounds 100% right.

I think the problem with a lot of people’s clones is that they use audio with background noise or poor mic quality. Either that, or they don’t use enough samples. Unfortunately, older actors have voices that fall under all three of these categories.

The best replications by far use ebook readings, podcast samples, or video game voice lines ripped from the game files.

Post
#1523041
Topic
The Rise of Skywalker: Ascendant (Released)
Time

Alright, here are the Kylo lines:
https://drive.google.com/file/d/1KSFAmimMyfzztgsI4oLv1G3pSmYl4o_J/view?usp=sharing
https://drive.google.com/file/d/1WZoNXMltOBa4shpRnLHZSBZSdggjuNwr/view?usp=sharing

Like I said before, Rey’s line after he speaks should be reinstated.

As for the “You were right”, “What Palpatine doesn’t know is we’ve become a dyad in the Force, Rey”, and “Captain Hux” lines, did those sound close to you, Hal?

I might not be able to generate Poe lines until this weekend because I have to manually string together all of his audio lines from the Lego TFA files into one large file.

Post
#1522996
Topic
The Rise of Skywalker: Ascendant (Released)
Time

So, Hal’s intention in removing “they sold you to protect you” was to imply that Rey’s parents instead paid for her to be hidden. Here’s a suitable replacement for that line:

https://youtu.be/QPWSB-_sSo4 (EDIT: New clip with improved wording)

Keep in mind that for this to be executed properly, we’d have to reinstate Rey saying “Stop talking!” after he says the final part of his new line. So the flow of the lines isn’t perfect because we’re actually missing a beat.

Post
#1522874
Topic
The Rise of Skywalker: Ascendant (Released)
Time

Yup, that’s the one. Also, here’s one reading of Sherlock’s altered line (I also have another one that delivers it a bit different):
https://drive.google.com/file/d/1SZujVo-iYqzoV6pO7maucZHMxXQhG0Ww/view?usp=sharing

And here is a possible Captain Hux for earlier in the film:
https://drive.google.com/file/d/14S44Xyd7X9JAxkTXO82X61ySvjyRZ0V5/view?usp=sharing

Post
#1522823
Topic
Deepfake Ideas - Index and Discussion
Time

I think that’s just unfortunately a side effect of the source material - her reading a book. Not a lot of opportunity to be whispering there. I think the first clip I posted managed to get her to somewhat whisper due to the way the AI had her deliver the line. But after regenerating other lines a bunch of times I haven’t been able to get a similar effect.

Post
#1522681
Topic
Deepfake Ideas - Index and Discussion
Time

NeverarGreat said:

I just stumbled upon this and was blown away. It uses this AI to create remarkably clear dialogue, and based on the spot-on pronunciation of the made-up words in that sample as well as some words which were clearly misread, this example was made by speaking the words to the AI and having the AI tune the voice to a specific character. If this is true, then it is a vast improvement over the FakeYou AI which failed to make the cut in my edit.

So what do you think? I’m thinking about a few choice discussions in the OT and how they could be transformed with a few edits from this AI.

I have a voice library for Carrie Fisher/Leia right now. I literally just put an entire 2 hours+ of her reading her book as input. Here is a clip I created:
https://youtu.be/-LzqBD8--H4