I Tested Google’s New AI-Powered Transcription App on IPhone, and It Was Surprisingly Accurate.
Google is once again introducing an AI-powered service—this time, an offline dictation program using the “Gemma” architecture. But instead of incorporating it into the Gemini app or making it a Gemini feature, the company has decided to release it as a standalone iPhone app with the catchy name ” Google AI Edge Eloquent .”
I decided to try the app on release day, although the privacy policy made me wary. Google states that your location, contacts, identifiers, device diagnostics, contact information, user content, usage data, and “other” data can be linked to you, while purchases and other diagnostic data may be collected but not linked to you. That’s a lot of data, especially for an app that advertises that “audio, private conversations, and personal data never leave your device,” and I’m not sure I would have wanted to download it if it weren’t for that. But as the saying goes, if the service is free, you are the product. I’ve reached out to Google for clarification and will update this article if I hear back.
How to try Google’s new AI-powered transcription app.
Once you download the app, setup is simple: you record a sample phrase the app will prompt you to say, then choose between “On-Device Mode,” which works entirely offline and saves your conversations online on your device; or “Enhanced Text Processing,” which saves the audio on your device but uses Gemini to “process” the text, which requires sending data to the cloud (and presumably is where all the privacy policy data mentioned above is sent). However, you won’t need to leave Gemini running for basic editing of your transcript—the app is designed to remove filler words like “uh.” Keep in mind that the app seems to open in “Enhanced Text Processing” mode by default—at least that’s what happened for me. But simply tapping the toggle in the upper-right corner of the main screen switches you to “On-Device Mode.”
I had some trouble getting the app to work: every time I tried to test it, it claimed I hadn’t spoken at all. But after pairing AirPods with my iPhone and then unpairing them, the app seemed to work. To test the app, I played the intro to this Audio University video on YouTube , which is entirely dialogue-based. Once the app worked, it immediately began transcribing the video with near-perfect accuracy—at least toward the end. I watched as the app entered incorrect words, then corrected and replaced them as subsequent words provided context. After finishing recording, the transcript was nearly identical to the video transcript, save for a few oddities: it mistook the phrase “If this is our first meeting” for “This is our first meeting” and recorded the same sentence twice. But otherwise, it’s a perfectly usable transcript of the beginning of the video.
Here, you have several options—especially if you invite Gemini to assist you. Right away, you can click the pencil icon above the transcript to edit it manually if you want to correct any errors the AI has polished. Above, you can view “Usage Statistics,” including the number of words spoken, words per minute, and the number of edits made by the AI. If you enable Gemini, you’ll have access to additional AI editing tools, including “Key Points,” “Formal,” “Concise,” and “Long.” Once you’re satisfied with the transcript, you can click the copy button to move the text to the clipboard and paste it elsewhere. In the “History” tab, you can review your previous transcripts and return to them for editing (either manually or with the AI). In the Dictionaries tab, you can add lesser-known words that you use frequently but that the AI might not recognize, which will improve the accuracy of your future entries.
During my brief testing, the app worked well, and I appreciate the ability to use it only on the device. I would definitely consider it as an alternative to iOS’s built-in transcription feature if it seemed faster or more accurate, especially given its more powerful features—assuming that using it on the device truly means protecting my data from Google.