OpenAI is finally rolling out its voice assistant to ChatGPT Plus subscribers but it’s missing some of the features the company showcased in its original demonstration.
Advanced Voice Mode (AVM) was first demonstrated in May but concerns over safety alignment and AVM’s ability to reproduce copyrighted content stalled its release. The controversy over one of the voices called “Sky” sounding a lot like Scarlett Johansson likely didn’t help either.
OpenAI humorously hinted at the frustratingly long wait for AVM in its announcement of the rollout.
AVM comes with five new voices, including one with an Australian accent, but fans of the movie “Her” will be disappointed to find that Sky’s voice has been removed. The new voices are Arbor, Maple, Sol, Spruce, and Vale, bringing the total to nine.
Another feature that users won’t get yet is AVM’s vision capability. When OpenAI originally demonstrated the voice assistant it could access the phone’s camera and respond to visual data. AVM only has the voice modality for now.
There’s no word on when, or if, AVM will get the gift of sight. It might simply be a strategic move to reduce the computing demands on OpenAI servers as users rush to try out AVM.
AVM does come with some new features though. AVM will store a user’s “custom instructions” for how they want the model to speak. You can specify a specific cadence or how you want it to address you. It also allows a user to provide personal background so the AVM can offer more contextually relevant responses.
The “memory” feature enables ChatGPT to remember conversations to reference in later interactions.
Users are already sharing videos of creative ways they’re using AVM.
If you’re in the UK, the EU, or Switzerland, you may have to wait a while before AVM is activated on your device. OpenAI hasn’t indicated when the feature will be rolled out globally.
