Google has sunk a lot of money into AI, and its success has been spotty. But the new Pixel Screenshots app is an example of when it works. It intelligently analyzes and organizes your screenshots. You’ll be able to find them later over simple conversational search cues, too. That’s a feature that goes to show Google is doing something right in applying AI.

Not all AI features, however, have proved successful. For instance, Gemini summaries in Google Recorder have been botched. On-device Gemini Nano promises real-time transcription summaries, but quite frequently, it doesn’t work and gives errors like “Transcript is too short” or “Transcript is too long”.

These are particularly notable problems with longer recordings. Summaries are most useful on the lengthier content, but these current limitations hinder this. At least in the case of setbacks like these, improvements are what the new Pixel 9 series is for, including a new Gemini Nano with multimodality that finally allows for longer transcripts.

Now, it can hold recordings up to 41 minutes long, which makes a big difference. It still sometimes fails at recordings over an hour. As of now, Gemini Nano in multimodality is only available on the Pixel 9 series and there has been no indication it will extend to older Pixel models.

While developments in AI from Google are admirable, such as Pixel Screenshots, the boundaries for Gemini Nano remain there. In the case of older models of Pixels, users may not get any advantages from fresh improvements.