> What is this? Uberduck.ai is a community-based synthesis website / community consisting of many voices. > What does it do? It can synthesize voices. Essentially a Text to speech. > What does it use to synthesize? We currently use Tacotron 2 & Talknet (aka singing section of the the site) to synthesize our voice models. > How could I try it out? Head to https://uberduck.ai/ > What is different to this than any other TTS service online? The voices are much more human-like, can express more emotion and overall is an upgrade to your classic Microsoft TTS voices. > How could I help contribute? There are many ways: - You can upload and submit your trained Tacotron 2 model to https://uberduck.ai/submit - You could help out a community member - You could spot bugs Anything more that, in general, helps the uberduck.ai community as a whole. (You also get a cool role!) > Are there any Google Colaboratory notebooks to start? Yes, they are located in #841437191073955920. They are maintained and updated. > How could I request voices, share datasets, or suggest features? You can gain access to those channels by agreeing to the guidelines in #842148452464853002 > The site isnt letting me make any voices say anything! The site may be experiencing high amounts of traffic. In the mean time, check in #842838618511114280 and or #825016537949011969 for updates. > Are you a copy of 15.ai or FakeYou? No; although similar, we do not compare others' endeavors to ours. > How would I make an AI voice?

> When ever I try to synthesize, I can't play the audio / it comes blank / gives an error! It may be because the server is receiving a lot of requests #842838618511114280 You may try to clear cookies and refresh, or retry the prompt. Another reason may be profanity in your prompt. > How could I receive highest priority for synthesizing to stop the audio errors? You may support us on Patreon to receive high-priority requests! (Server boosting does NOT count) > How can I request a voice to add to the site? You may do so by agreeing the guidelines in #842148452464853002 first, then head to #841461499946860576 and do /voicesuggest > Why do I need to sign in to use uberduck.ai? The main reason is the server being overloaded with requests. Another reason is that we are implementing account features periodically. > How can I submit my pre-trained model onto the site? You can do so here: https://uberduck.ai/submit > Its not letting me sign in! It says "Error" If it says an error, try making sure that you deleted your cookies, then refresh the site and try again. If that doesn't work, then log into discord in a different tab, then in another tab try logging into uberduck.ai. > Why was this voice removed? The voice was probably removed due do false information being spread, and we chose to take it down. Otherwise, the original voice actor / person asked for it to be taken down.

> Why does the person sound like its gibberish / having a stroke? This is because some models were not trained on longer transcripts. This could end up as gibberish if the prompt is long, so a good way to prevent this is to split your prompt by sentence, and input each at a time, later combining it together in a audio editor. Like audacity or even in a video editor. > Is there an all-in-one guide to start? Thanks to @User for this guide: https://awesomeface349.gitbook.io/uberduck-written-tutorial/

> How do you use the bot? Go to #842056330835656745 and use the bot command, /quack-help, in that channel for a list of voice commands to generate TTS clips with the bot

> Is there a way to make the TTS voices speak in emotionally expressive ways, e.g. angry, sad, and/or happy? As of now, no, not really. Custom emotional expressiveness cannot be done with the model architecture technology that Uberduck uses. Perhaps it can/will be doable in the future, though

> How do you suggest voices? Go to #842148452464853002 and press . Next, in #842056330835656745, use the /voicesuggest bot command.

> Why isn't every voice in the 🎵 side on the website? The reason is simple: Singing models are trained with TalkNet, whereas speaking models are trained with Tacotron 2. And we don't have TalkNet models of every voice that is in the normal speech synthesis section on the website.

> When will my voice model be on the website? I sent it to uberduck sometime ago. We have a big backlog of voice models that we haven't put on the website yet, and because there are so many model submissions that have been sent to us, we don't get to all of them, unfortunately. You may contact us directly or in a ticket channel that you can open in this server in cases that you want to get particular models to be put on the website. Although, there is no guarantee that we would fulfill those requests quickly. 😦

> I use uberduck to learn English. Do you think the engine is suitable for this use? If you're interested in using uberduck to learn English, message zwf directly to talk with him in regard to that. As of now, most of uberduck's models don't enunciate all the way correctly, but this is a use case that we would love to support and encourage.

> Why does the @Uberduck Agent bot fail and give me an error message that reads "This interaction failed." when I use commands with it? The bot is kind of broken. It works sometimes but not the majority of the time.

> Why are some of the Latin American Spanish voice models not on the website? Because the owner of those voice models has requested them not to be on uberduck.ai due to trolls using them to make the voices speak racist stuff en Español. The voice models in question were Bob Esponja , Drake, and Toretto.

> The outputs that some of the models synthesize sound robotically hoarse. Why is that? Those models haven't been trained with enough ||good|| voice data. ⬅️ This applies to Tacotron 2 voice models, TalkNet voice models, and HiFi-GAN neural vocoder models.

> How do I invite the @Uberduck bot to my server? Go to uberduck.ai, scroll down, and click/press the "Add TTS to Discord" link at the bottom of the page.

> What is the difference between

SpongeBob SquarePants (post-sequel)

,

SpongeBob SquarePants (pre-movie)

, and

SpongeBob SquarePants (regular)

? The "post-sequel" SpongeBob voice model has been trained with voice data from the "modern" era of the franchise, AKA 2015⁠–⁠present. The "pre-movie" SpongeBob voice model has been trained with voice data from Season 2. The "regular" SpongeBob voice model has been trained with voice data from multiple SpongeBob SquarePants video games, especially those that were released in the post-movie era.

> What are

CRUST

and

AITCH

? CRUST is a multispeaker pipeline model (not to be confused with hybrid voice model) that has been trained with data of 168 voices. More importantly, it is a base model on which a pipeline model can be trained & fine-tuned. Although, it has now been superseded by AITCH, which is a multispeaker pipeline model that has been trained with data of 1,000 voices.

> How many voice models do I have to submit to obtain the "contributors" role in this server? I've submitted [this many] voice models to Uberduck. It isn't a matter of how many voice models you have to submit to Uberduck to be given the role. Obviously it isn't as though the role is given automatically, so it is your responsibility to let us know that you've submitted voice models to Uberduck.

> You're an admin of Uberduck, so aren't you supposed to know how to solve [this problem]? Please don't expect/assume that we know how to solve every problem just because of our role in taking charge of Uberduck. I mean, just because we're admins of the service doesn't mean we know everything about it.

> Can you put my voice model on the website? I sent it to uberduck sometime ago. It helps to know what the model is so that we can put it on the website, silly goose.

> The ability to login with my Discord account is gone. How do I get around this? You have to sign in via the Magic Link method now. If you enter the email address that's associated with your uberduck account, you'll get this in your Inbox. Hit "Log In" and then bam, you're instantly logged back in.

> Why have all of the public reference audio clips been taken off the website?

> Where is the list of forbidden voices? Have you checked up your butt? Here it is: https://docs.google.com/spreadsheets/d/1n6KZpcahyckjmzgsH90ghgNs5pcEQ57baavbjAI_ed8

> Isn't it too early to celebrate April Fools day at the moment? Not for people who are in the UK.

> Isn't it too early to celebrate April Fools day at the moment? Not for people who are in the UK.

> The Uberduck Discord bot is broken in terms of voice channel stuff. Yes, but the maintainers of the bot don't have time to fix it yet.