Using ChatGPT with Voices and Images

   
Image from ChatGPT website

A new major update on OpenAI's ChatGPT model came out a week ago. This update enables users to interact with the model in more ways than just using text. These ways are namely speech and sending images. In addition, ChatGPT can respond to the user's voice with its own voice or send pictures to the user if requested. These features have been implemented by a variety of third-party extensions but it has finally been brought natively into ChatGPT. This advancement has been anticipated by the community for over a year now.

Users can now talk to ChatGPT via a microphone and the model responds with voice using text-to-speech technology and it is acclaimed as a good quality one. This essentially means that people can now talk to ChatGPT like talking to a real person. This also opens a way for people with certain disabilities to use ChatGPT more easily. Another advantage of that would be the effectiveness of using ChatGPT, people can interact with the model faster and understand the response more easily with the audio support on top of text.

Users can also upload images from their devices to ChatGPT and ask for outputs from the model. ChatGPT can also respond with images using Dall-E 3. Although the image inputting may be sloppy sometimes, it could be very useful for enhancing the context comprehension of ChatGPT where the user may find it hard to fully explain the situation with words. Like the voice inputs, the visual inputs can also increase accessibility. 

One interesting example given by the developers that utilize both of these functionalities in an accessibility sense is a blind person taking a photo of a pair of his/her t-shirts and asking ChatGPT which one is red and which one is blue. Although this is just one example, it is foreseeable that this would be useful in many cases.

There is one issue though. All these features have been added to GPT-4 and this version of ChatGPT is not free of charge. Today, users have to pay a monthly fee of $20 to gain access to GPT-4 from its website. The default plan only covers GPT-3.5 which has fewer features including the voices and images. 

For those who do not want to pay this much money on GPT-4 whether because they just want to try and see it, or they straight up don't think that it's worth it, there is actually a way to access GPT-4 features. Users can access GPT-4 for free by using Microsoft's Bing search engine. Although its use and reliability are limited compared to the ChatGPT website, it is a great way of trying out the new features and doing test runs for free. It should be noted that text-to-speech is not available at Bing right now.

I definitely recommend checking out GPT-4 for everyone. Go toy around and see what interesting ideas you can come up with. It is very inspiring.

Thanks for reading!

Oğuz Arslan

Comments

Post a Comment

Popular posts from this blog

Pizzas in 10 Minutes Ontology Tutorial

Image
Protege is a no-cost, open-source platform that offers tools for building domain models and creating knowledge-based applications using ontologies. It is supported by a growing user community.  This blog will explain the standard 10 Minute-Pizza ontology. You will learn how to use the W3C Web Ontology Language (OWL) by following this tutorial. Note that this is not a complete walkthrough and it does not explain everything step by step. For that, you can check out the Pizzas in 10 Minutes page on the Protege Wiki website. Although the tutorial is slightly outdated with the most recent version of the program, it should not be hard to go through the steps.  In the beginning, having a basic structure defined on paper similar to the image below should be useful.      Image taken from  OWLMap: Fully Automatic Mapping of Ontology into Relational Database Schema To use Protege, you should first download Protege Desktop , install the Matrix and Annotation templ...
Read more

Why we hate each other on the internet? (by Kurzgesagt)

Image
Last week, the popular YouTube science channel Kurzgesagt uploaded a video titled "The Internet is Worse Than Ever – Now What?", talking about how divisions in opinions on the Internet have become bigger and scarier. The Internet is Worse Than Ever – Now What? Kurzgesagt (word meaning "in a nutshell" in German) is a science channel that creates content based on many research topics including physics, biology, psychology, politics, etc. Kurzgesagt has 9 different YouTube channels where the same videos are published with different language narration. The biggest one, with English narration, has 21.5 million subscribers. Kurzgesagt logo Kurzgesagt is a very well-respected channel in the YouTube science community. This is not only due to its clean animations and impactful narration but also because they do very intense research for every video while being extremely transparent about where they gathered the information. Under every video's description section, they ...
Read more

Gephi Quick Start Guide

Image
Gephi is an open-source network analysis and visualization software often used by researchers and professionals to explore and analyze complex networks. It's trendy among those working in social network analysis, biology, transportation, and more fields. Gephi allows users to import network data, visualize it in various layouts, and apply different algorithms to analyze the structure and properties of the network. Users can also customize the appearance of nodes and edges, making it a powerful tool for creating visually appealing network visualizations. Gephi Logo This blog will cover a quick starter guide for Gephi. The official tutorial can be found in this link . Our version follows similar steps but in an updated version.  First, you need to download Gephi from this link  and install it on your computer. This guide is done with the Gephi 0.10.1 version for Windows. The tutorial uses the LesMiserables  sample dataset.  After starting Gephi, you need to import LesM...
Read more