OpenAI CEO Sam Altman says the first users will start to get access to GPT-4o Advanced Voice in the next couple of weeks, but this will be a limited "alpha" rollout.
The company is testing the full capabilities of GPT-4o, a new type of Omni model released during its Spring Update in May. Unlike GPT-4, this natively multimodal model can understand speech directly without converting it into text.
This makes GPT-4o both faster and significantly more accurate when acting in the role of voice assistant, even allowing it to pick up on tone and vocal intonations during a conversation.
Users have been waiting patiently for access, but OpenAI says safety testing must be completed first. Some have briefly gained access, and there have been multiple demos of its capabilities, but most users won’t get it until later this year.
What is GPT4o Advanced Voice
alpha starts later this month, GA will come a bit afterJuly 18, 2024
GPT-4o Advanced Voice is an entirely new type of voice assistant, similar to but larger than the recently unveiled French model Moshi, which argued with me over a story.
In demos of the model, we’ve seen GPT-4o Advanced Voice create custom character voices, generate sound effects while telling a story and even act as a live translator.
This native speech ability is a significant step in creating more natural AI assistants. In the future, it will also come with live vision abilities, allowing the AI to see what you see.
Other use cases for Advance Voice include having it act as a very patient language teacher, able to correct you directly on pronunciation and help improve your accent.
“ChatGPT’s advanced Voice Mode can understand and respond with emotions and non-verbal cues, moving us closer to real-time, natural conversations with AI. Our mission is to bring these new experiences to you thoughtfully,” OpenAI said in a statement last month.
Why the delay in launching GPT-4o Advanced Voice?
OpenAI is one of the most cautious artificial intelligence labs, taking significant time to security test, verify and put guardrails in place for any new major model.
Altman has also called for regulation of frontier-style models like the upcoming GPT-5 or world models like Sora due to the risk they present to society. This caution has allowed other companies to begin to catch up with OpenAI, and GPT-4 is no longer the only top-tier model.
The company was concerned that GPT-4o Advanced Voice, without appropriate guardrails, could offer potentially harmful information or be used unexpectedly. To tackle this, they’re gradually releasing it to trusted users first and then more widely over time.
“As part of our iterative deployment strategy, we'll start the alpha with a small group of users to gather feedback and expand based on what we learn,” a spokesperson explained.
“We are planning for all Plus users to have access in the fall. Exact timelines depend on meeting our high safety and reliability bar. We are also working on rolling out the new video and screen sharing capabilities we demoed separately, and will keep you posted on that timeline.”
More from Tom's Guide
- I just tried Runway’s new AI voiceover tool — and it’s way more natural sounding than I expected
- Hume AI brings its creepy emotional AI chatbot to iPhone
- ChatGPT Voice could change storytelling forever — new video shows it creating custom character voices
