On April 6, with no press release or marketing campaign, Google quietly dropped an app on the iOS App Store: "AI Edge Eloquent". A voice dictation tool that works 100% offline, automatically removes filler words, and turns speech into polished text without needing an internet connection. For LATAM professionals who produce content, write documents, or simply want to write faster, this represents a concrete change in their daily workflow. It is also Google's first major move in local (on-device) AI for personal productivity, and it anticipates where AI is heading in the coming years.

What AI Edge Eloquent is and what sets it apart from Siri or Google Dictation

AI Edge Eloquent is a standalone Google app available on iOS that combines voice recognition with on-device language models (Gemma models, Google's open source family) to offer high-quality dictation that solves the weak points of current alternatives.

Its main features:

• It works without an internet connection: the processing happens entirely on the device

• It removes filler words ("uh", "um", "well", "like") in real time, without user intervention

• It punctuates and automatically formats the text (paragraphs, commas, periods) based on the context

• It supports multiple languages including Latin American Spanish with regional variants

• It exports directly to Notes, WhatsApp, Gmail, and Google Docs

What sets it apart from the most used alternatives:

• vs. Siri Dictation: Siri requires an internet connection to process voice in most cases. AI Edge Eloquent works completely offline.

• vs. native Google Dictation: the basic transcription is functional but does not remove filler words or apply intelligent formatting. AI Edge Eloquent does.

• vs. Whisper (OpenAI): Whisper is high quality but requires technical configuration to run locally. AI Edge Eloquent is a ready-to-use app with no setup.

How it works: on-device Gemma models and why local AI matters

The heart of AI Edge Eloquent is Google's Gemma models (the open source version of Gemini), optimized to run on mid-to-high-end mobile hardware without draining the battery or requiring constant streaming to cloud servers.

The three critical advantages of on-device AI over cloud models:

1. Privacy: the audio and text never leave the device. There is no recording on external servers. For professionals who dictate sensitive information (contracts, strategies, customer data), this is a fundamental differentiator.

2. Speed: minimal latency because there is no round trip to a server. The transcription appears in real time without the perceptible delay of cloud solutions.

3. Availability: it works without internet, on the subway, in areas with poor signal, on trips abroad, or in any context where connectivity is not guaranteed.

For Google, this is strategic beyond the app itself: it demonstrates that Gemma models are capable enough to run on consumer devices, and it sets the technical precedent for integrating local AI into future Pixel and Chrome OS devices. According to The Next Web, Google is reportedly planning to expand the family of on-device Gemma models to multiple app categories over the next 18 months.

Concrete use cases for LATAM professionals

Who benefits most from AI Edge Eloquent in the context of everyday work in the region?

• Content creators: dictating scripts, ideas, and captions on transport or while walking is significantly faster than typing. A creator can capture a complete idea in the time it would take to write the first paragraph.

• Journalists and writers: capturing ideas and quotes from interviews without depending on a recorder or subsequent manual transcription.

• Sales professionals: dictating CRM notes, follow-up emails, and meeting summaries immediately after a call, while the context is fresh.

• Students: taking notes by voice during classes or readings and getting structured text directly, without the filter of typing speed.

• Designers and creatives: capturing briefs and concepts quickly without interrupting the flow of visual thinking.

The common denominator: any professional who needs to turn ideas into text quickly, cleanly, and without depending on an internet connection.

Comparison: AI Edge Eloquent vs. the market alternatives

The only alternative with comparable local processing capabilities is OpenAI's Whisper, but it requires technical configuration to run on-device. AI Edge Eloquent has the advantage of being a ready-to-use app with no additional setup.

According to TechCrunch, Google's quiet launch on iOS (without Android yet) suggests a controlled-market testing strategy before a broader rollout.

If you're interested in exploring this topic further, you can also read generative AI tools for marketing and content in LATAM.

Recommended Coderhouse courses

To understand how the AI behind tools like AI Edge Eloquent works and learn to use them to create content and work more efficiently:

• Content Creation with AI Course: learn to generate scripts, texts, and multimedia content with AI. Ideal for professionals who want to produce faster by integrating tools like AI Edge Eloquent into their workflow.

• AI Engineering Course: technical foundations to understand how models like on-device Gemma are built and deployed. For those who want not only to use but to understand the AI behind the apps.

• Introduction to Artificial Intelligence Course: the starting point to understand concepts like on-device models, generative AI, and how these technologies will transform work in the coming years.

Frequently asked questions

Is AI Edge Eloquent available on Android?

At the time of launch, it is only available for iOS. TechCrunch reported that the Android version is in development, with a launch expected for the second half of the year. Google did not confirm an exact date.

Is it really free?

Yes, the app is downloaded for free from the App Store with no in-app purchases at the time of launch. Google did not communicate monetization plans, although it is likely that in the future it will offer premium features or integration with Google One.

How accurate is the dictation in Latin American Spanish?

The first user reports indicate an accuracy of over 95% for Latin American Spanish with a neutral accent, with good results for Argentine, Mexican, and Colombian accents. The Gemma models were trained with multilingual data that includes regional variants of Spanish. Performance may vary with very marked accents or very technical and specialized vocabulary.

Is the audio I record saved on Google's servers?

No, according to the technical documentation published by Google. The processing is entirely on-device and the audio is not transmitted to any external server. However, as with any app, it is advisable to review the terms of use and the privacy policy before dictating sensitive or confidential information, since the conditions may be updated.

Does it work on older iPhones?

The app requires iOS 17 or higher and is optimized for iPhone 13 or newer, since the chips of those generations have the computing capacity needed to run Gemma models in real time. On older models there may be higher battery consumption and greater latency in the transcription.