Moshi AI Voice Assistant Launched by Kyutai Labs as GPT-4o Rival

Kyutai Labs on Wednesday launched Moshi AI, an artificial intelligence (AI) chatbot that responds verbally in real time. The French AI company announced that Moshi’s entire audio language model was developed in-house. It can also modulate the voice to express emotions and respond in different speaking styles. The AI ​​model is freely available to the public. Currently, the AI ​​model limits conversations to five minutes. Interestingly, OpenAI has also announced similar voice capabilities with the release of GPT-4o, but it has yet to be released.

Moshi AI Features

Company states that the AI ​​model was developed in six months with a team of eight people. During the unveiling of the AI ​​model at an event in Paris, Kyutai Labs said that Moshi is not an AI assistant, but a prototype that can be used to develop tools for different use cases. It also made the chatbot public hereUsers can enter their email address and join the queue, but Gadgets 360 staff were able to access the platform immediately without any wait time.

The platform’s interface is quite minimalistic. There’s a simplified AI design where users can check the loudness of their voice when speaking. There’s a text box where only the AI’s responses appear. Another box at the top displays technical details like audio duration, latency, and audio misses.

At the very top is a button to disconnect the call. Currently, the maximum call duration is five minutes. The description page emphasizes that Moshi can think, speak and listen simultaneously to maximize the flow of the conversation.

Gadgets 360 found that latency is extremely low and the AI ​​often responds instantly. However, there are a few cases where the response time lag can exceed 10-15 seconds. But this could be due to heavy server load. However, sometimes the verbal prompts would not register at all, even after three-quarters of the volume meter was filled.

Moshi AI interface
Photo Credit: Kyutai Labs

Gadgets 360 also found that the AI ​​model can respond with an emotional voice, and can speak in different styles and using different voice modulations. The AI ​​model is also connected to the internet and can retrieve answers to the questions that require searching the internet. Interestingly, the chatbot does not allow for text prompts and speech is the only medium to communicate with it.

Kyutai Labs has stated that the AI ​​model will be open source. However, the AI ​​company has yet to host the model weights and code on a portal. Once it is available, users will be able to download and install it locally and it can be run on a disconnected device.