ChatGPT‘s Advanced Voice Mode arrived on Tuesday for a select few OpenAI subscribers chosen to be part of the highly anticipated feature’s alpha release.
The feature was first announced back in May. It is designed to do away with the conventional text-based context window and instead converse using natural, spoken words, delivered in a lifelike manner. It works in a variety of regional accents and languages. According to OpenAI, Advanced Voice, “offers more natural, real-time conversations, allows you to interrupt anytime, and senses and responds to your emotions.”
There are some limitations to what users can ask Voice Mode to do. The system will speak in one of four preset voices and is not capable of impersonating other people’s voices — either individuals or public figures.
In fact, the feature will outright block outputs that differ from the four presets. What’s more, the system will not generate copyrighted audio or generate music. So of course, the first thing someone did was to have it beatbox.
Advanced Voice as a B-boy
Yo ChatGPT Advanced Voice beatboxes pic.twitter.com/yYgXzHRhkS
— Ethan Sutin (@EthanSutin) July 30, 2024
Alpha user Ethan Sutin posted a thread to X (formerly Twitter) showing a number of Advanced Voice’s responses, including the one above where the AI reels off a short “birthday rap” and then proceeds to beatbox. You can actually hear the AI digitally breathe in between beats.
Advanced Voice as a storyteller
This is awesome actually
I did not expect the ominous sounds https://t.co/SgEPi5Bd3K pic.twitter.com/DnK8AVdWjV
— kesku (@yoimnotkesku) July 30, 2024
While Advanced Voice is prohibited from creating songs wholesale, it can generate background sound effects for the bedtime stories it recites.
In the example above from Kesku, the AI adds well-timed crashes and slams to its tale of rogue cyborg after being asked to, “Tell me an exciting action thriller story with sci-fi elements and create atmosphere by making appropriate noises of the things happening (e.g: A storm howling loudly)”.
look on OpenAI’s works ye mighty and despair!
this is most wild one. You can really feel like a director guiding a Shakespearean actor! pic.twitter.com/GUQ1z8rjIL
— Ethan Sutin (@EthanSutin) July 31, 2024
The AI is also capable of creating realistic characters on the spot, as Sutin’s example above demonstrates.
Advanced Voice as an emotive speaker
Khan!!!!!! pic.twitter.com/xQ8NdEojSX
— Ethan Sutin (@EthanSutin) July 30, 2024
The new feature sounds so lifelike in part because it is capable of emoting as a human would. In the example above, Ethan Sutin recreates the famous Star Trek II scene. In the two examples below, user Cristiano Giardina compels the AI to speak in different tones and different languages.
ChatGPT Advanced Voice Mode speaking Japanese (excitedly) pic.twitter.com/YDL2olQSN8
— Cristiano Giardina (@CrisGiardina) July 31, 2024
ChatGPT Advanced Voice Mode speaking Armenian (regular, excited, angry) pic.twitter.com/SKm73lExdX
— Cristiano Giardina (@CrisGiardina) July 31, 2024
Advanced Voice as an animal lover
— Ethan Sutin (@EthanSutin) July 30, 2024
The AI’s vocal talents don’t stop at humans languages. In the example above, Advanced Voice is told to make cat sounds, and does so with unerring accuracy.
Trying #ChatGPT’s new Advanced Voice Mode that just got released in Alpha. It feels like face-timing a super knowledgeable friend, which in this case was super helpful — reassuring us with our new kitten. It can answer questions in real-time and use the camera as input too! pic.twitter.com/Xx0HCAc4To
— Manuel Sainsily (@ManuVision) July 30, 2024
In addition to sounding like a cat, users can pepper the AI with questions about their biological feline friends and receive personalized tips and advice in real time.
Advanced Voice as a real-time translator
Real-Time Japanese translation using #ChatGPT’s new advanced voice mode + vision alpha! Yet another useful example! pic.twitter.com/wDXrgYQkZE
— Manuel Sainsily (@ManuVision) July 31, 2024
Advanced Voice can also leverage you device’s camera to aid in its translation efforts. In the example above, user Manuel Sainsily points his phone at a GameBoy Advanced running a Japanese-language version of a Pokémon game, and has the AI read the onscreen dialog as he plays.
The company notes that video and screen sharing won’t be part of the alpha release but will be available at a later date. OpenAI plans to expand the alpha release to additional Plus subscribers “over the next few weeks” and will bring it to all Plus users “in the fall.”