Google Introduces Chirp 3, AI Model for Voice Generation

Google has announced the addition of “Chirp 3”, its speech-to-text and HD text-to-speech model, to the Vertex AI platform. 

Chirp 3 was announced On March 17, 2025, at the “Gemini for the United Kingdom” event in London,

Key Points

• Chirp 3 is an audio generation model that adds a range of custom voices with human-like inflections and intonations. 
• Chirp 3 will be available on Vertex AI starting next week, joining other frontier models on the platform such as Gemini, Imagen, and Veo. 
• Chirp 3’s HD Voices feature supports 31 languages, providing 248 unique voices with eight speaker options.
• Chirp 3 can be used for tasks like voice annotation, meeting transcription, audiobooks, podcasts, and building AI voice assistants.
• Google will implement usage restrictions to prevent misuse, with its safety team currently reviewing guidelines, according to Google Cloud CEO Thomas Kurian.

Chirp 3 Vs. Whisper and Polly: A Battle of Speech AI Models

Google’s Chirp 3 competes with leading speech AI models like OpenAI’s Whisper, Microsoft’s Azure Speech, and Amazon Polly.

While Whisper focuses on accurate multilingual speech recognition, Azure Speech offers real-time transcription and translation.

Amazon Polly is known for generating lifelike voice outputs.

Chirp 3 stands out with its HD Voices feature, supporting 31 languages and 248 customizable voices.

Its human-like intonations and versatile applications, from voice assistants to podcasts, give it an edge in realistic audio generation.

News Gist

Google announced Chirp 3, its speech-to-text and HD text-to-speech model, for Vertex AI.

Supporting 31 languages and 248 voices, it enables tasks like transcription, podcasts, and voice assistants, with safety measures in place to prevent misuse.