Connect with us

Softwares

This is Tele-Prompt, a new AI that helps you intervene in conversations with clever phrases and ideas

Avatar of Thomas Grimm

Published

on

1672308510 840 560 jpg

A young developer named Daniel Gross, who has already worked for some time on Apple’s staff, along with his colleague Natt Friedman, they just announced a new Artificial Intelligence that gives you ideas of what you can say in a conversation. The AI ​​is called Tele-Prompt and also does not need an Internet connection since it is developed privately within the computer.

It is not the first AI created to help us in a conversation. ChatGPT has recently taken the world by storm thanks to its capabilities, but Tele-Prompt has very particular characteristics and can be used both in a work meeting and in an informal conversation.

ZAO, the Chinese MOBILE APP that through DEEPFAKE turns you into DICAPRIO in SECONDS

Speech Recognition and Python


Now let’s see how it works. First, as we already discussed, “everything happens privately on the computer” and the code that is shared on GitHub. The first tool necessary is a voice recognition system that captures what is said.

This AI includes a Python framework for sentence embeddings known as a SentenceTransformer. This framework can be used to compute phrase or text embeddings in more than 100 languages.

Five tools to write texts with artificial intelligence and save time

For now, Tele-Prompt is in the experimental phase. This is to say that it looks very “green” in the examples shared in a Twitter video by developer Daniel Gross. Even so, it is one more example of how the Artificial Intelligence advances at great speed to reach the whole world and cover different aspects. In the video you can see that sometimes the program can give obvious answers.

For example, one of the people in a meeting asks for a solution to a problem and the AI ​​says that “the solution will come when the people in charge are responsible of the problem”.

Or phrases that can be bombastic or ingenious such as “freedom means that you define yourself as you want to be defined” or “Care is what matters, not things or status, just people sharing care and joy” or “Freedom is a state of mind.”

Softwares

This AI that allows you to clone your voice (and anyone else’s) is filling the Internet with celebrities saying stupid things: this is how you can use it

Avatar of Thomas Grimm

Published

on

1675261259 840 560

An AI that allows anyone, and in a matter of seconds, to clone the voice of any other person (or, of course, oneself) is beginning to take the Internet by storm. The controversial but popular 4chan forum, without going any further, has been flooded with Audio deepfakes of celebrities like Emma Watson or Joe Rogan filled with violent or pornographic comments.

And all since, this past weekend, access was opened publicly to the ElevenLabs platform, perhaps one of the most significant examples of how far AI-based Technologys dedicated to audio cloning and synthesis have advanced.

Apart from the undeniable realism of its voice synthesis (light years away from canned voices like Loquendo, so popular years ago), the company itself boasts that the audio clips it generates require little or no editingwhich encourages its use in tasks such as dubbing or audiobook production:

When video deepfakes emerged, we all thought that an era of massive misinformation was inaugurated… and in the end its main application was the generation of porn. It remains to be seen what will happen to their audio equivalents

ZAO, the Chinese MOBILE APP that through DEEPFAKE turns you into DICAPRIO in SECONDS

Somebody put order here

Of course, the fact that it does not present none of the safeguards we are used to finding in imaging services to avoid creating certain types of content ‘sensitive’ and/or based on real people.

In a thread posted on Twitter last Monday, ElevenLabs recognized that some of its users were abusing its application: “[observamos] a growing number of cases of misuse of voice cloning”, and warned that they are able to “trace any generated audio back to the user”.

Likewise, it has put on the table the possibility of implementing protections such as the User identity verification or manual approval of each request voice cloning. However, its use remains unlimited for now.

Advertisement

we have tried it

When we register for free on the platform, we are granted access to a series of tools. First of all, we will find the simple voice synthesizer that allows us to convert texts to sound files using, in principle, predefined voices in American English.

However, if we make use of the ‘Voice Cloning’ function, we can upload audio with a voice recording to the platform (“it must last more than 1 minute and not contain background noise”, clarifies the website), which will allow us to create a personalized voice from which we can return to the synthesizer and create an audio based on our voice… or anyone else’s.

screenshot 39

Voice Cloning Step 1

screenshot 40

Step 2

That yes: the service is, for now, optimized for the American English accent; so it is possible that, when generating the audio, you do not finish recognizing yourself in the resulting file.

screenshot 41

The classics never die. Although it sounds horrible in American, let me tell you.

That has been our case, after generating a personalized voice based on that of a server: reading Don Quixote in a perfect Kansas accent (up state, down state) has not finished sounding ‘very me’, but US users claim not to suffer from this problem.

The free account only allows create up to five custom voices and synthesize to audio 10,000 characters of text per monthIn addition, the resulting audios are subject to a non-commercial license. There are, from there, several payment options, the cheapest being $22 per month.

Via | The Verge

Image | Based on an original by Jason Rosewell on Unsplash

Advertisement
Continue Reading

Softwares

The Spanish-speaking audience does not stop growing on Twitch. But we’re still so underrated

Avatar of Thomas Grimm

Published

on

1675258875 840 560

It is undeniable that Twitch has undergone a great boom in recent years accompanied by the strategies that They have followed different content creators who have attracted a large audience to the platform. But the problem that resides is the great differences that can exist between the English and Spanish-speaking communities, mainly due to the differences in investment by advertisers.

But the great reality is that despite this difference in investment and prestige, the truth is that the Spanish-speaking community is one of the most powerful in terms of growth in the number of viewers and also in number of creators. And this is something that was recently evidenced at the ESLAND Awards where the community gave a big blow to the table.

Getting started on TWITCH How to set up OBS and PERSONALIZE your CHANNEL

The Spanish-speaking community is growing at a fast pace

One of the examples that Bloomberg has chosen to express this large Spanish-speaking community is that of Rivers. A streamer who studied engineering in Mexico in 2020 when she started playing Call of Duty on Twitch motivated by her friends. And as they went live, more and more people began to see this girl who wondered if she should spend more time being a streamer and less time studying. Obviously, this is something that she did not like in her family, although in the end it seems that she made the right decision to currently have a live audience of 19,000 viewers on average.


And this is motivated by a substantial increase in interest in the transmission of the Spanish-speaking public. Specifically, between 2019 and 2022, the audience in Spanish multiplied, going from the seventh category to being very close to English in a second position, as Bloomberg has detailed.

These are all the winners of the ESLAND 2023 awards: Ibai conquers Twitch and is proclaimed 'streamer of the year' for the second time

But it is not something that is coincidental, but rather it is due precisely to the fact that Latin America has about 80% of people between the ages of 10 and 65 play video games. But the problem in this case is that Latin America does not receive much attention from investors, and it is that it is at the bottom when it comes to investment in digital advertising. It is even believed that this market is considered ‘incipient’. Specifically, the country as a whole is located behind Africa and the Middle East.

Income

Given this boom, from Twitch they have increased their presence in the region with more staff and adding marketing teams, creative partners and sales. But this is something that can be very sad, since currently in the record for viewers on the entire platform, Spanish streamers stand out with Ibai and its Evening of the Year II, and also the successes that many creators in Latin America are reaping. This confirms that Spanish is booming on the social network, although it is not being given the recognition it deserves.

Source | Bloomberg

Advertisement
Continue Reading

Softwares

Chronophoto is the new addictive online game that asks us to guess what year a photo was taken: that’s how it works

Avatar of Thomas Grimm

Published

on

1675254272 840 560

Last year Wordle surprised us with an addictive word game that captivated millions of people to enter the web every day for a new challenge. After its success, a multitude of clones and games with similar mechanics arose, and the New York Times benefited from the success of the original by acquiring it and adding it to its list of hobbies.

Chronophoto, for its part, proposes a completely different idea, although it is reminiscent of the one mentioned above due to its addictive formula and simple mechanics. And it is that it is a free web game in which all we have to do is guess what year the picture was taken that they present to us

A simple game to guess the year the photos were taken

The game is tremendously simple, and that’s where its popularity comes from. Chronophoto will show you five photos in which you will have to guess what date they were taken. For this you have a bar that you can move from the year 1900 to 2022.


In each photo you will be able to zoom up to a limit, so you can take advantage of elements such as people’s clothing, vehicles or various objects to try to get closer to the correct date. Although there are photographs of more or less historical moments, you will also find day-to-day images, so it will be in the latter where you have to benefit even more from the elements that appear in the photo.

Framed is the game inspired by Wordle to find out movies that will make you become a movie buff

The closer you get to the year the photo was taken, the more points you’ll get, while if you get too far away, you won’t get any points. The maximum score is acquired when you nail the year, thus obtaining a total of 1,000 points.

Chrono4

The points will accumulate as you progress through the round, and at the end you will get a summary with all the points earned.

The best thing about Chronophoto (and the worst) is that once the round of five photos is finished you can start over with different images. Be very careful with this, because we already warn you that it is a truly addictive game.

Unknown creator and a pending authorship issue

At the moment the author (or authors) remains in the shadows, although you can contact them in case you think there is an error in the date of any image. You just have to click on the ‘About’ button.

Advertisement

Another issue that has been discussed in forums is about the intellectual property of the photographs that appear in the game. And it is that according to some point out, they could be images stolen from the internet. This is a problem, since the page does not give credit to the original authors of each photograph and it is surely an issue that must be solved immediately.

At the moment Chronophoto It is completely free and does not include advertising some on the web. Of course, the web does not save the score obtained, so you will have to keep your maximum score to try to beat it the next time you play.

Continue Reading

Trending