Google launches AI Edge Eloquent app for polished speech to text conversion

Google has launched a sophisticated new application designed to transform raw, disorganised speech into polished professional text.

The "Google AI Edge Eloquent" app, currently available exclusively on iOS, utilises the company's Gemma-based automatic speech recognition (ASR) models to process audio directly on a user's smartphone.

Unlike traditional transcription tools, this application is specifically engineered to maintain the speaker's intended meaning while automatically removing filler words such as "um" and "ugh," as well as mid-sentence corrections.

The resulting output is a crisp, accurate piece of text that can be further refined into formal prose or condensed into key bullet points.

A primary feature of the Eloquent app is its commitment to user privacy. By running machine learning models locally on the device, the app ensures that audio files, transcriptions, and personal data do not leave the smartphone.

This on-device processing also allows for full offline support, making it a secure option for sensitive dictation. While certain advanced features can utilise Google Cloud for enhanced processing, users retain control via a toggle button on the record screen.

Additionally, the app allows for the creation of custom dictionaries and can integrate with Gmail accounts to import frequently used vocabulary.

The application is currently free to use with no usage restrictions and is compatible with mid-range hardware.

While an Android version has not yet been announced, Google has confirmed it is evaluating other platforms, including desktop versions for tasks such as coding and document dictation.

Google was founded in 1998 and remains a global leader in artificial intelligence and search technology.