TTS Software?

[jan]

Hi -

I posted this in the 'Help With Files' section first, but this is a better place to ask, I think:

I went looking for software to convert a text file into a voice. The best I found, which was not good, was an IBM Watson application that produced so-so files.

What TTS software do wwm people use? Where do you all get the voices (I like sexy UK girl voices myself? Does your software allow you to adjust pronunciation of tricky phrases?

I'd like to write some of my own stuff, but I need some decent TTS voices and sound edition software to get started, so any help is appreciated.

I see Natural Reader, which starts to cost money once you want extra voices, and LinguaTec, which seems to be $49 per voice, without publishing permission, or a cool $499 for the pro/publish edition.

Thanks for any suggestions,

jan

[Fizbin]

For sound editing, Audacity is pretty much the go-to software.

For speech, Cepstral and Ivona have pretty good voices, and if you have a later version of Windows some speech is built-in, though the voice selection is limited without buying from 3rd party sources. Hazel is the (quite decent) built-in UK English female voice (you need to install the UK English language files to get her). If you download the free Speech SDK, you also get a player as sample code (it comes compiled)

[jan]

Thanks for taking the time to reply.

I never heard of Cepstral, and I'll look into it. I did play around with Ivona before, and will try them again - and with whatever Windows can offer.

Thanks Again,

jan

[Leeiah]

There is lots of free stuff out there hun

[daugo]

if you want something that sounds more human find a translated copy of vocaloid. though this would make your files have a heavy japanese accent, I have tried it and it does sound very engrish. heard the newest version of vocaloid sounds better though.

[shrinksarah]

Here is a link to a free online TTS website. I've tried it with a few scripts and it works well.
http://www.fromtexttospeech.com/

[kzarin2]

Amazon polly and learn the few tags for ssml.

You just need the prosody, breaths and speak tags tbh.

I use Amy, British female but there's of choices.

Its a very good free speech to text service based on amazon aws and bests everything else I've seen.

[Maleko7]

I will recommend Polly too. Although it was a bit of a hassle to figure out how to do it (and get the thing to do emphasis and other tricks), I was impressed with the outcome.

I use the free subscription for development.

[lauraparidae]

I use Lua to generate markup for the "say" command in MacOS, which can be used to create an AIFF file that is then compressed with ffmpeg. I could use a shell script to mix in additional audio tracks or make multiple audio tracks and then mix those together somehow. I like this approach because I can iterate on the script, sometimes just changing one word or a few letters and I don't have to do any repetitive work at all to generate the audio file.

The markup / template system I used in Lua can do some neat tricks, but the TTS controls are currently very simple. I recently found a web site that allows you to do TTS within your web browser and the voice markup commands are quite similar: http://webhypnotist.pw

One advantage of generating the whole script from a markup & template system in a real programming language is that I can change the underlying code and generate TTS markup for some other engine... For instance Google Wavenet voices sound almost perfect and I could convert my file to use that engine, if I wanted to. Despite sounding very real, there was an edge in the voices that wasn't as soothing as the slightly more unnatural voices in Macintalk.

Other advantages of the system are that the scripts are modular and configurable. I already have four versions of the script I uploaded here recently (two of them private custom versions just for myself). So, if I work on a new script and improve an induction or deepener that I used on another script, that other script will also improve with no extra effort.

Please comment and review the file I uploaded - I'm interested in feedback and further development. Sometimes I worry that I'm relatively alone in the world with my transformation kink, but when I found this site a few days ago, I immediately felt that this could be the right place for the file I made. (I hope you agree.)

[slondonfun]

Balabolka I find good.

[Maleko7]

I found a great way to do TTS but it may be over your head.

Amazon has TTS but you have to sign up as a developer. I don't remember all the details but can come up with detailed instructions if his is something you want to peruse.

I take a script and send it through their system and download the MP3. I then delete it off their servers. There are something like 6 voices and the results are pretty good. Oh, did I mention this is free?

I then load the MP3 into Audacity if I want to massage the audio a bit. Sometimes I add a track or change the pitch to make my voice a bit less nasal and deeper.

Also they have a markup language where you can add pauses rather than rely upon Audacity.

Here is "Matthew" reading the above. One issue. He says Google because I created the audio before I remembered that it was an Amazon process not Google. OOPS

By the way, the link I am posting is a free service to store things and share them using blockchain. Little parts of this file are stores all over the place and assembled at the time you click the link. For reasonably small files there is no charge.

The audio file sample: https://app.blackhole.run/#jpPUmCtdrh1C ... 11J11GWfqg

Information about Blackhole file transfer: https://www.producthunt.com/posts/black ... e-transfer

[Maleko7]

I found Amazons instructions.

https://aws.amazon.com/polly/

I use the free tier and have never paid for converting to MP3. Then again, I am not exceeding 1 million character a month

[otmen]

Personally, if it’s a file I don’t care about I use text to wav and the Windows default voices. I speaking is a problem and you really want that real human timbre then 15.ai (which is the IRL) has some pretty impressive stuff with fully adjustable inflection but I find it’s better to just put the text in and look at the different option it generates and chose the best one. The Narrator from Stanley Parable is an awesome voice, so is the turret from portal but I haven’t used that one a ton so it may be a bit untrained since it only has 2 minutes of original audio to train on (although was one of the better voices when I tried it). You could also hire someone to voice act for your files, just put an ad up online or a job board and be honest about the job and you might be surprised by the people with natural talent. You should also generally keep up with new AI TTS (and yes I know this is from 2016 and it’s now 2021) as AI is really heading to new frontiers in terms of realistic recreations of audio, video, text, and so much more. I use an AI to help me write sort of like auto correct and it generates some pretty good stuff sometimes, others it gets a little off topic but it still works amazingly good.

[panis]

You could install Bluestacks, download the Acapela TTS Voices app, buy a voice on the cheap, then record it through an audio recording program. Voices sound very realistic and can be adjusted.

[lauraparidae]

TTS has come a long way in the past few months. Right now Elevenlabs.io is going viral and I think for a good reason. Has anyone made files with the new AI voices? The only one I have played with is Tortoise, but I'll probably try Elevenlabs soon.

[mistschaufel]

I find this site to work well for me:

https://www.naturalreaders.com/online/

[cloudstalker]

Has anybody been working with AI voices to make hypnotic scripts? Any you recommend?

[naughtywish14]

Wanted to add voicemaker.in to this list with the audio book creation pack these voices sound so real and i think its a pretty good value as well

[harleenzaftig]

I've been playing around with Elevenlabs(dot)io

You can even clone your voice if you opt to pay for a subscription!

I've been messing around with using my own voice with other peoples' scripts to see if they put me deeper.

[geekdh]

I was doing some stuff a year ago with creating video ads to run on YouTube. I used something call “Mic-Monster” and was pretty satisfied with the results. There’s a free version (limited in text size and number of voices) and a “Pro” version for $50 a month. The pro version has about 20 or 30 different voices (American and British, and male and female). (All information here is from memory and may be outdated.

Hope that helps.

Geek

[geekdh]

I was doing some stuff a year ago with creating video ads to run on YouTube. I used something call “Mic-Monster” and was pretty satisfied with the results. There’s a free version (limited in text size and number of voices) and a “Pro” version for $50 a month. The pro version has about 20 or 30 different voices (American and British, and male and female). (All information here is from memory and may be outdated.

Hope that helps.

Geek

[slutinmyhead]

I recently tried AI voice cloning through Eleven Labs and while there were some where the pacing changed or there was distortion, I'd say about 80% came out amazing. You have to pay for it, but it's not crazy expensive. Personally, I like recording. It relaxes me and I like playing with my inflections from file to file, but my partner does it more as a favor to me and this was great for getting several in her voice without her having to do anything. Eleven Labs isn't the only game in town either. And while I haven't done this yet, I do think I could probably have it rerecord the parts that ran into problems and just cut and paste using Audacity. Recording those parts myself to splice in didn't work well because you can hear the difference just from echoes in the room, background noise, etc.