The BEST Text to Speech Software | This is WILD
HTML-код
- Опубликовано: 26 сен 2024
- Guys, I think I finally found my favorite text to speech software! Descript Overdub turns your own voice in AI generated text to speech! In this video, you'll see how Descript Overdub works, and why I think it's the best text to speech software I've found.
If you want to try Descript Overdub for yourself, click here: get.descript.c...
If you want to see my Synthesia Custom Avatar, you can see it here! • I got DEEPFAKED! | Syn...
Visit www.jennjager.com to learn more.
Shop My Gear!
Camera: amzn.to/330jcFP
Lights: amzn.to/2G5Ztvm
Microphone: amzn.to/30kyeFc
Headphones: amzn.to/2RQhOPF
Music Library: bit.ly/3raqXTU
For event bookings, sponsorships or collaborations email fan@myvideo101.com
For professional video production services, visit www.plumproduct...
Follow me on Instagram @jennjagervideo
Follow me on Facebook: / myvideo101
When you purchase products through links on my page, I may receive a small commission at no additional cost to you.
If you want to see my Synthesia Custom Avatar, you can see it here! ruclips.net/video/9YrYF0wpisg/видео.html
it seems like that initial exclamation point made your entire overdub sound like you were frustrated at the listener. haha. but it's still pretty automagical
Hi Jenn, thanks for sharing. I immediately heard the difference between your real voice and the A.I. from the first clip. It was a bit too fast the way it read the sentences and maybe if there was a way to control the speed of the voice would make it much more accurate. What do you think? As we know naturally when we speak, speed, pitch and tone make a huge difference how your speech is perceived.
What are you worried about this is how it is in the world deal with it
Do I have to use my own voice or can I use AI(not copy of my voice) and still get monetized?
I can hear a big difference in the timbre of the voice as if they were recorded in different environments. As far as telling whether it was you or not is another story. That is quite amazing.
I am legally blind and cannot use the text editor in descript at all. The overdub is so amazing I will spend an extra hour going to a text editor copying pasting into descript until it gets to be correct lol
Descript is fantastic! My wife thought I was talking live through a whole script! She thought it was pretty scary!
We are certainly living in the future! If you didn't tell me this was being overdubbed by Descript I would have totally believed everything was you. This is the thing about technology, it keeps getting incrementally better and as time progresses it seems like magic. I await to see the full AI version of you in the next video!
This is very cool! I'm a filmmaker. We do a thing called ADR, where we bring the actor in to overdub their voice when there's too much background noise on location, or for some other reason the audio recorded on set doesn't work. This could save a lot of time and $$$. Thanks for sharing!
be careful using this for that service, last thing you'd want is for them to own at least in part, the voices in your film... Seek a commercial license instead
I can definitely tell when it is you speaking and when it was the program. It sounds like the same person speaking either way, but you speaking sounds enthusiastic and the program sounds stern. It wouldn't matter for a short overdub of a segment, but you wouldn't want to use it for the whole presentation. At least until they add some emotional markup to it.
A couple of things that I have noticed with several of the Ai text to speech programs is that they speak to fast so trying to get them to slow down can be a bit of a problem, and also trying to convey emotions can be impossible to do. But given time these text to speech programs will only get way much better.
How much better?
ruclips.net/video/MT_u9Rurrqg/видео.html
@jennjager Descript Overdub does a correct American speech pattern with a flat or downward inflection at the end of the sentence indicating it's a statement. The upward inflection is reserved for a question. The reason you hear it differently is because your speech patterns fall into the uptalk/upspeak with the upward inflection at the end of a definitive statement sentence. If you pay attention and by eliminating the uptalk/upspeak and record it again then the Overdub will make a better match. You can search the topic of uptalk/upspeak in RUclips you'll find many videos explaining why it should be corrected. All the best.
Your inflection is sweeter and somewhat not too confident (sometimes it sounds like you're asking a question when you are actually making an affirmation) while the AI is more authoritative and assertive. It almost sounds as if you are angry. lol It's the 5% that breaks the magic, but it's almost there. Very fun stuff.
Awesome! That is absolutely your voice. I was at another window just hearing your video and at first I didn't even noticed it was the TTS that said "Let's hear what this voice sounds like".
I'm hoping they include this feature in Brazilian Portuguese, they are already transcribing to my language, would be awesome to have my own voice in my native language!
the replacement over the corrected audio sounds angrier than the context.
The A I. voice sounds more aggressive than yours. You are right about the app not being very simple to navigate in. I tried it a week ago, I wanted to generate a voice (not mine) but I didn't manage to do that. I wrote my piece and no sound came out of it.
Like an evil clone
I just imagine how dangerous this is. What would happen if someone trains the model with the voice of a politician/leader or any other popular personality. Then create some content that would trigger political unrest among public /communities
We don’t give a fuck anymore what they say due to the excessive lying and misinformation they spew.
What does matter what ai is made to say as fake speech?
It won’t be any worse than what the say already
Its also possible to immitate video with deep fake.
I am pretty sure it is happening now anyway with similar software!
It is a real issue that we are likely to see soon. There will also then be people denying they said things.
It will have a negative impact because people already fall for easily debunked misinformation on things like vaccines, so it's only a matter of time before fringe groups pick it up.
@@goodgame3374 very true
Great video, it was really insightful, keep them coming.....
wow! I've recently been diagnosed with a carcinoma tumor on my vocal cords. I'm going to radiation treatment but I'd love to have my voice recorded/sampled so that in case I lose my voice completely I can use future tech to talk to people! thank you!
Still sounds robotic but who knows how this will sound in the future with further ai developments.
I agree. Just from the time i've listened to you, the generated version is just a little flat..almost slightly mechanical compared to your actual voice. But i think that me watching this video beforehand probably accenuated it more in my mind. You could likely do any video you have, not let the person ever have seen your content, and they wouldn't notice anything.
Pretty wild stuff, I ran into your video when I found out that the voice of Darth Vader in Obi-Wan Kenobi was not James Earl Jones, it was all AI. Great video explaining this process.
I really love the way the human Jenn speaks and explains. For me there is no real personality, no enthusiasm in the generated voice.
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, Such a nice option isn't it!
Maybe it's because I was an audio engineer, but I could tell right away. That inflection is part of your personality coming through when you talk, and the software doesn't match it at all. However, for short mistake replacements, it would likely be fine. I wouldn't use it for a long project, though.
Interesting software and testing! Thanks for sharing.
It occurs to me that people who face a change in their voice, such as those who have throat cancer or famous actors who want to continue doing voiceovers after age changes their voice, would benefit by establishing their voice in this app early in life. If it's acceptable, think of creating podcasts with only a script. There's one possible catch, though; for years, we've been able to type text and have the computer read it back in its own voice, but there are words the computer will not pronounce correctly. Usually, there's a technique to add marks and commands to force the correct pronunciation of difficult words and my question here is to ask if there's a way to change the pronunciation of a word. Thanks, Jenn. A great review as always. I want it!
naive asf
I compressed the descript audio, and EQ a bit. So the amplitude is bigger, and its sounds more organic.
How do you do that?
@@sumdomguoy548 I tried it using GarageBand. It’s mainly used for music but you can use it to compress audio and mess with the Eq
It sounds like your voice, but the inflection is no better than any of those apps. I would definitely not pay $ 24 / month for something like that. Maybe a one time purchase of the app would be reasonable, but this seems like a rip-off to me.
Jenn..I saw your video first time. I must say that I enjoyed that bubbly voice of yours and the so relatable expressions... its so refreshing. You will do outstandingly well. . You are underrated Dear. I am seeing 140k subs today. Hope the next time I see you will be crossing a million. Rock !!
I can immediately tell it's not you speaking (robotic) but it's a so-so stand in as your natural diction is more precise and clear than most. I can see it working best for voices that are a certain quality but may not be close for other voices.
This is wild! I downloaded descript by mistake and stumbled across this video. But now I'm super excited to use it!! Lol
I could hear the slight difference, but that is crazy how good it was.
I can DEFINITELY hear the difference. The text to speech part sounds a little like you have a cold. But it is pretty close.
I can tell the difference. Needs some pacing and inflection tweaks. Not the best TTS I've heard. Not the worst TTS I've heard. Definite progress from 2 years ago.
Very clever Jenn, indeed! Thank you so much! 🙏🔝🔝🔝
It sounds like older text to speech I've heard but with (pretty much) your voice, which is cool I guess. Just not a replacement as it still sounds quite robotic, unless manual tweaking is powerful enough without being time consuming.
Wow superb. I never knew the descript voice was playing. I thought the whole voice was your voice only.
This was so cute! I really need to resume going out by myself soon and feel that main character energy.💜
Experts in what is AI and what is the real person's voice will make millions as trial experts.
I think you’re right
I heard a *huge* difference between your voice and the Descript program. Firstly, the inflection is...staccato and hard on the first syllable. It also sounds perpetually angry, and I just don't get that vibe from you 😂. If you use the training sentences they provide to train the program on your voice, it might give a much different result.
If you just record a bunch of actual spoken word audio in descript using the in programs recorder or by uploading a file, if you apply your overdub speaker label to the audio, you can then select a few words in a sentence and right click and goto the overdub drop down menu and create a new voice style, then your over dubs using that style should fit in with the audio in that sentence you apply it to better. You also get to keep the style, and if you think your overdubs just sound better with that style just use it instead of the default. I find one of the styles I've made just sounds more natural then the one they sent back to me.
This tool seems so amazing and fun to use!!!
Thanks, Jenn for making this video, thank you RUclips for showing it to me! ❤
Totally awesome... probably, some future software can even "learn" with more samples given to even analyze inflection , pauses.. etc.
I think it would have gotten much better results if you read that training speech if you aimed to read it with a more perky "let-me-entertain-you" kind of voice rather than a let-me-pronounce-each-word-with crystal-clear enunciation type of voice
Sounds great, can it also be used with languages other than English?😀
Cool! Watching from the Philippines and just came across your channel. After searching through your video topics, I subscribed, as I appreciate the options and learning experience.
Great review yet again! I agree. Descript is certainly a game changer. I use it quite a lot and it is hard to tell the difference when using the overdub feature. I am looking forward to when the development team overhaul the UI to make it less clunky and more userfriendly. Keep up the good work!
Not saying simp things like "game changer" would be a game changer at this point.
Great idea and video. Ii I were watching a video where this technique has been used I wouldn't be able to tell the difference
I prefer to use Colbass
great video! Really like the way you interact with the viewer, and was very helpful!
Wow! That sounds scaring accurate! 😅
We’ll look back on this time as the breakthrough for the singularity. Where we really picked up momentum for the end of humanity. It’s been fun, goodbye
Thx for including this amazing product. I wonder if I put a transcript of another language, do I have to record my own voice again in other language?
The staccato annunciation is a little much. Would be cool if you could somehow soften the consonants. The AI sounds like you're almost spitting out the consonants.
sounds real, no uncanny valley effect, how did it match up with the original video clip being a bit shorter
ok I'm convinced! Thank you! Great video, I'm scurrrred LOL.
We should never believe anything we see or hear again. Unless it’s in person
Great work, thanks!
Fascinating technology! Although I was quite disappointed when I read on the company´s website that this service is only availabe in English for now.
LOL! your thumbnail!! 😂 Poor girl!!🤣
Your voice is amazing! I don't think the AI voice does it justice... Just my opinion
Wow. Super impressive & definitely a game changer.
I think the difference is only noticeable when the original is present for compairison
I think this is amazing and have been purchased the pro to work out this program... My problem is how do you make the voice change for the styles for the voice we made ? I'm lost..... Yes the GUI is a bit confusing..
Thanks for this helpful video! "Descript Overdub" does sound really good!
I heard the difference between your real voice and the AI generated voice but only because I was focused on the sound of your voice, not paying attention to what you were actually saying (if that makes sense). But if you had inserted a clip where else in the video with tell us, I don't think I would have noticed.
Hi Jenn, you are my favorite for learning. Thanks to you, I have Descript Pro. I have a video that's been published on RUclips. Now I want to use my AI voice to dub over my regular voice on the video, (I keep getting errors). do you have a vid for that? Or reference me to one? Many Thanks. Bill T
Excellent video. First time I hear of such a technology.
3:20 what's stopping someone from training an AI on your videos and then signing such a form with a recorded message?
Is there a way to edit the inflection or emphasis of words with this software?
Could you create character voices and select them for audiobooks? I can do many voices, so wonder how that would work with this.
You've done an outstanding review of this software. Thank you for sharing
Well, I could certainly tell the difference between your natural speaking cadence and the overdub. As the producer of a sci-fi themed rock-oriented music radio show, I work with AI-generated speech a LOT! My entire cast of the Starfighter Centaur on my program "Zombies of the Stratosphere" is computer generated. The biggest hurdle I have encountered is trying to get the nuance and cadence correct for my crew members. This involves lots of punctuation edits like adding -or eliminating-dashes, commas, quotes, and periods in ways that differ from "normal" script writing meant to be read aloud. Adding prosody cues helps to generate some speech nuances like raised excitement levels, or stupefaction (think-when a cast member's body is inhabited by an alien life force), or even sickness. Most often, in order to rectify the problematic section of the script, I end up rewriting the sentence over and over, or abandoning it entirely to take a different approach. If you watch many RUclips videos critically you can plainly hear that the voice on so many of them, while pleasant, simply says a lot of things in a wonky manner. In addition to proper cadence, appropriate emphasis at key points in the script is EVERYTHING! Even when, like your example, it is in the producer's own AI voice. However, given the rapid progress of AI "art forms" in general I am sure that it will not be too long before your natural looking and sounding avatar can continue making video presentations well into your 60s while still looking young, vibrant and attractive as you are now! To hear my humble efforts using AI-generated speech you can hear my radio program here. My cast comes in at the beginning, about 20 minutes after, 40 after, and at the end: kpov-od.streamguys.us/Zombies%20of%20the%20Stratosphere_stream.mp3
Good insights!
Thanks for showing this I’m a hypnotherapist and I’ve lost my voice with laryngitis multiple times so it’s now mostly croaky so I’ve delayed creating hypnosis audios. I’m hoping this has natural sounding tones and not robot if i try to use it for recordings or videos 😊
Why the ending is so wild?! If it's the best text to speech software, then where are the SSML tags?
that's insane! let's hear what this voice sounds like!
I m looking for a way to build my own using libraries instead of submitting my voice for others to train the model on and send me back the result
Thanks for this great video, and you are right about the awful and unintuitive UI 😖 It took me a while even after reading documentation that in order to access the stock voices, I needed to go into writing mode, and from there I could then click on a "Speaker Label" to access the voices. It just wasn't self-explanatory enough.
But the stock voices are amazing.
I don't have a great Overdub voice yet...I want to try to use one of their default voices for my next video just to test... :) Thanks Jenn!
How do we know the audio n this video is not the generated voice?
very useful and well explained, thank you!
Where I want to delete someone's voice & replace it with text to speech, I'd rather enter length of time of original segment to be deleted rather than use trial & error with a speed slider like Text 2 Speech android app, etc, etc. This is particularly necessary to get a victim voice off of a long crime video.
At 3:35 when you sent your voice to the AI to be trained, then suddenly your clothes were changed. This is a super AI not only change your voice but and your clothes too. Wow....
7:44 I wAnted to work with, there I noticed some autotuning-like mishap.
It did sound like you! Wow!
Fantastic presenting! Brilliant! 👋
It's probably less robotic sounding if you use their training scripts. The inflection is way off but that's probably because 10 minutes is the recommended minimum, and that's from the script that was designed to catch all the nuances. 10 minutes from a video recording probably isn't enough time.
I don't know if I'm more impressed or scared by this...
If you want your videos monitized then do not use text to speech. RUclips is demonitizing videos with it.
I really dislike billing structure of these services. "its $12/month but we charge for the year." they should advertise what they're going to bill not an irrelevant breakdown. that's like goin to Starbucks and paying $12 for a 12 oz coffee and they "justify" it by breaking it down as its only $1oz. but you'll have to buy a minimum 12 oz, whether you want 12 or not. smh.
What if you are dyslexic and just want to read text online? Can this program read for you?
Try speechify
@@michaelmerrifield5914 Thank you for replying ! :) I already tried that one!
Great content! Thanks for sharing.
This descript overdub voice TTS is the best thing out there! Truly amazing! But it would have caution Don't like I did try uploading 9 hours of your video for it to train on your voice lol this is still too new and it will just break everything lol Little bits at a time :-)
the longer you train your voice on descript the better it is.
would you check out this tts REVOICER please?
How can I solve this? Each time I extract an audio voice in a video and translate it to another audio voice and when I try to put it back to the video it does not match the video any more. Is either the voice is moving faster than the video or the video is moving faster than the voice please how can i make it match to look more real?
Mind. Blown.wowza. can't wait for next weeks vid.
Glad you're excited!
yeah yeah this is a very sophisticated program or application🤠🤠🤠🤠🤠
sorry, but you sound WAAAYYY better than the computer generated version
This is scary and amazing at the same time!
Is there a way to add voice inflection?
I think I prefer the overdub voice. It seems sexier! 😍
can i make it into conversation between 2 people?