OpenAI’s New Voice AI Can Listen, Think, and Talk Back in 70+ Languages

OpenAI has just unveiled a trio of advanced audio models within its Realtime API, marking a significant leap forward for voice-powered applications. These models—GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper—collectively push voice AI beyond simple question-and-answer exchanges. Instead, they enable systems that can truly understand a speaker, take meaningful action, and sustain a natural, flowing conversation. If the company’s demonstration is any indication, we are witnessing the next evolutionary step in how voice AI models function.

This new OpenAI voice AI suite offers developers unprecedented capabilities. But what does each model bring to the table, and how might they reshape everyday interactions? Let’s break it down.

GPT-Realtime-2: Smarter Conversations with Context

The flagship model, GPT-Realtime-2, integrates GPT-5-class reasoning into live voice interactions. This means it can tackle complex requests without losing the thread of a discussion. For example, it can call multiple tools simultaneously—like checking your calendar while searching for a nearby restaurant—and even narrate its actions aloud with phrases such as “let me look into that.”

Building on this, the model boasts a larger context window of 128K tokens. As a result, interactions feel more coherent over extended sessions. Developers also have the flexibility to adjust reasoning effort based on the complexity of a request, making it suitable for both simple commands and intricate tasks.

Real-Time Translation Across 70+ Languages

Perhaps the most exciting addition is GPT-Realtime-Translate. In many ways, it brings us closer to a universal translator akin to what we see in science fiction. The model supports live speech translation across more than 70 input languages and 13 output languages. During the demo, even when a new participant joined and spoke a different language, the system seamlessly translated both speakers into English in real time.

This real-time translation capability is a game-changer for international communication. Whether you’re in a business meeting with global colleagues or traveling abroad, this AI can bridge language gaps instantly. For developers, it opens doors to building apps that facilitate cross-cultural conversations without delay.

Streaming Transcription with GPT-Realtime-Whisper

Finally, there’s GPT-Realtime-Whisper, a streaming transcription model. Unlike traditional speech-to-text systems that wait for a speaker to finish before delivering the full transcript, this model converts speech to text as the speaker talks. This makes it ideal for live captions, real-time meeting notes, or any voice-powered workflow where waiting is not an option.

Moreover, its efficiency means that applications like live event subtitling or instant dictation can become more accurate and responsive. Developers can integrate this into tools that require immediate text output, enhancing user experience significantly.

Who Can Use These Voice AI Models?

Currently, OpenAI has released these models exclusively for developers. However, the apps they build will inevitably affect everyday users. For instance, a developer could create a real-time translator app that lets people converse across languages effortlessly. Several companies are already testing these models. Zillow is building a voice assistant that searches homes and schedules tours from a single spoken request. Priceline can check flights and hotels, cancel bookings, and make new ones. Vimeo is using the technology for real-time transcription, among other use cases.

Pricing starts at $0.017 per minute for Whisper, $0.034 per minute for Translate, and $32 per 1M audio input tokens for GPT-Realtime-2. This tiered structure allows developers to choose based on their needs, from basic transcription to advanced reasoning.

The Future of Voice AI

As these OpenAI voice AI models become more widely adopted, we can expect a shift in how we interact with technology. Voice interfaces will move from being mere novelties to essential tools for productivity and communication. For more insights on AI advancements, check out our article on AI Trends for 2025. Additionally, learn how to build voice-powered applications with these new tools.

In conclusion, OpenAI’s latest release represents a major step forward in making voice AI more intuitive, responsive, and multilingual. Whether it’s for business, travel, or daily tasks, these models promise to make conversations with machines feel more human than ever.

Perplexity Personal Computer Mac App Goes Wide: Meet the Autonomous AI Agent for Your Desktop

Perplexity has officially launched its Personal Computer Mac app for all users, and this release marks a significant step forward in how artificial intelligence can interact with your local machine. Instead of just answering questions in a chat window, this AI agent works directly on your Mac — searching files, operating inside native apps, browsing the web, and connecting to over 400 tools — all without constant human supervision.

You give it a task, and it gets to work. You only step in when it needs your approval. It’s similar in spirit to Claude Cowork, but Perplexity’s approach is purpose-built for deep integration with macOS.

What Can Perplexity Personal Computer Do for You?

If you have ever lost an hour jumping between apps, hunting for a file, or manually pulling information from different sources, this tool is built for exactly that frustration. It can work across your local files, native Mac apps, connected tools, and the web — all in one go.

Here is a fun example Perplexity showcased: open Notes, press both CMD keys, and ask it to handle your to-do list. It figures out each task and gets them done. You can also ask it to clean up a messy folder and sort everything into properly named project folders.

Building on this, the best part is that you stay in control. It checks in when a decision matters, and every action it takes is reversible. Think of it as a capable assistant — not an unsupervised one.

What’s the Best Way to Run Perplexity Personal Computer?

According to Perplexity, a desktop Mac is the ideal environment. Since the entry cost of a Mac mini is relatively low, it’s the recommended option. Running the AI agent on a Mac mini allows it to operate continuously, even when you are away from your desk.

You can kick off a task from your iPhone, and the agents keep working on your Mac back home. You can also approve requests from any device and let it run overnight for complex projects.

However, the Mac mini is currently in short supply, so you might have to wait or grab any available model. The new Perplexity Personal Computer Mac app is available today for all users. Everyone gets everyday features like search, attachments, and dictation.

It is not yet on the App Store, so you will need to download it directly from Perplexity’s website. Learn more tips for using Perplexity AI effectively to get the most out of this tool.

How Autonomous AI Agents Are Changing Productivity

This launch reflects a broader shift in the AI landscape. Instead of passive chatbots, we now have autonomous AI agents that can act on your behalf. Perplexity’s approach is particularly notable because it works locally on your Mac, meaning your data stays on your device rather than being processed entirely in the cloud.

For power users, this opens up possibilities like automated research, file organization, and cross-app workflows that previously required complex scripting or third-party automation tools.

Practical Use Cases for Perplexity Personal Computer

• Research and summarization: Ask it to gather information from multiple sources and compile a report.
• File management: Request a cleanup of your Downloads folder into organized project directories.
• Cross-app workflows: Have it pull data from a spreadsheet, create a chart in Keynote, and email the result.

Getting Started with Perplexity Personal Computer

To begin, head to Perplexity’s website and download the Mac app. Once installed, you can start giving it tasks immediately. The interface is straightforward: type or speak your request, and the agent takes over.

Remember that you can approve or reject actions at any time. This safety mechanism ensures you remain the decision-maker for critical operations. Explore other Mac productivity apps that complement this tool.

Final Thoughts on Perplexity Personal Computer

Perplexity has delivered a genuinely useful AI agent that feels like a natural extension of your Mac. It is not a gimmick — it solves real problems around multitasking and information management. As the technology matures, we can expect even deeper integration with macOS and third-party services.

Download it today and see how much time you can save by letting an AI agent handle the busywork.

AI chatbots have transformed how we discuss personal topics, including some of life’s most difficult moments. But this openness comes with responsibility. OpenAI is stepping up with a new feature called ChatGPT Trusted Contact, designed to bring a human into the loop when conversations take a serious turn.

Rolling out now for adult users, this optional setting lets you designate one person who can be alerted if the AI detects potential self-harm concerns. It’s a proactive move that blends technology with human oversight.

How Does ChatGPT Trusted Contact Work?

Setting up a ChatGPT Trusted Contact is straightforward but comes with clear rules. The person you choose must be at least 18 years old—or 19 in South Korea. Once you nominate someone, they receive an invitation explaining their role. They have one week to accept before the feature activates. If they decline, you can pick another contact.

The alert process isn’t automatic. When ChatGPT’s systems flag a conversation as concerning, the chatbot first informs you that your contact may be notified. It also suggests conversation starters to help you reach out directly. A small team of specially trained human reviewers then evaluates the situation. Only if they confirm a serious risk does your contact get notified—via email, text, or in-app alert.

What Information Is Shared?

Importantly, the alert doesn’t share chat transcripts or details. It simply states that self-harm came up in a potentially concerning way and asks the contact to check in. OpenAI aims to complete this human review within one hour, ensuring timely support without compromising privacy.

Why Is OpenAI Adding This Now?

This feature builds on earlier safety measures. Previously, OpenAI introduced alerts for parents when linked teen accounts show distress. ChatGPT Trusted Contact extends that protection to adults. It was developed with input from clinicians, researchers, and mental health organizations, including the American Psychological Association.

However, this feature isn’t a replacement for professional help. ChatGPT will still direct users to crisis hotlines and emergency services when needed. You can remove or change your trusted contact anytime, and contacts can opt out whenever they wish.

As AI becomes a confidant for many, ChatGPT Trusted Contact acknowledges that technology has limits. It’s a step toward blending digital support with real human connection. For more on AI safety, check out our guide on AI safety tips for users.

What This Means for Users

The reality is that people use ChatGPT for deeply personal conversations, whether OpenAI planned for it or not. Adding a feature like this is a move in the right direction. It also admits that a chatbot can only do so much.

If you’re considering setting up a trusted contact, remember it’s optional but potentially life-saving. For more on mental health resources, visit our mental health support page.

In summary, ChatGPT Trusted Contact represents a thoughtful evolution in AI safety. It combines automated detection with human judgment, offering a safety net without overstepping privacy boundaries.