ChatGPT GPT-4o New Feature Rundown: Voice Translation, Desktop Access, and Memory Controls

This update ties voice, images, and memory together around GPT-4o, turning ChatGPT from “something you can chat with” into “something you can use on the fly.” Below, we break down ChatGPT’s new features by the most common scenarios.

GPT-4o merges text, images, and audio into a single conversation

GPT-4o is positioned as “omni” (all-purpose). For ChatGPT, the most noticeable change is smoother multimodality: within the same conversation you can type text and also upload images and files, letting ChatGPT read the content directly and then reason about it, rather than only providing surface-level descriptions.

If you’re used to using ChatGPT to organize materials, this integration noticeably cuts steps: screenshots, spreadsheets, and PDFs no longer need to be converted to plain text first—you can drop them straight into ChatGPT to extract key points, compare differences, or generate checklists, lowering the communication overhead.

Advanced Voice and Real-Time Translation: Use ChatGPT as a portable interpreter

ChatGPT’s voice interaction feels more like a normal conversation: you can revise your request while speaking, and ChatGPT can respond more quickly without needing you to wait for it to “finish thinking” after every sentence. When you mix different languages in a conversation, ChatGPT supports fast switching and can provide near real-time, interpreter-style translation.

For people who often attend international meetings, you can have ChatGPT restate the same sentence in different tones, or translate spoken language into a more formal email version. For learners, you can ask ChatGPT to correct pronunciation approaches, provide synonym substitutions and example sentences, and practice more smoothly.

The desktop app is easier to summon, and chat history is easier to find

The value of the ChatGPT desktop app is “instant access”: on Mac you can use a shortcut (such as Option + Space) to open ChatGPT quickly, reducing the interruption of switching back and forth to the browser. The desktop app is also better suited for dragging in emails, screenshots, and files, letting ChatGPT take over organizing and rewriting directly.

Another practical tweak is chat search: once you have lots of conversations, scrolling through history is painful; now you can use keywords to locate old chats in ChatGPT, making it more efficient to continue asking follow-ups or reuse prompts.

Now that Memory is live, first learn “what to have ChatGPT remember”

ChatGPT’s memory is no longer just “the context of this single conversation”—it can continue to apply in future chats. Officially, memory is split into two parts: “saved memories” that you explicitly ask it to remember, and “chat history insights” distilled from past conversations, helping ChatGPT better match your preferences.

Even more important is control: you can turn off memory-related options in settings, ask ChatGPT to forget a specific piece of information, or use “temporary chat” for conversations that are not written into history and do not update memory. It’s recommended to let ChatGPT remember reusable preferences (tone, formatting, commonly used tools), while handling sensitive information in temporary chats.