Argmax Pro SDK for Android

• Argmax Pro SDK for Apple runs on millions of devices, powering real-time speech recognition with top-tier models such as Nvidia Parakeet and OpenAI Whisper.
• Today, we are introducing Argmax Pro SDK for Android in collaboration with Google.
• This Kotlin-first SDK brings Argmax's top-tier accuracy and real-time speed from Apple to Android, enjoying seamless NPU and GPU acceleration by Google LiteRT.

‍

‍

‍

Why LiteRT?

Our first foray into Android was in collaboration with Qualcomm. We deployed OpenAI Whisper on Qualcomm Snapdragon NPUs for speech-to-text on pre-recorded audio. This was built on the Google TFLite inference runtime, predecessor to LiteRT.

This was a promising start but there were several limitations to this approach:

• TFLite only supported Qualcomm Snapdragon NPUs. While Qualcomm builds industry-leading chips and has a significant market share, it is far from ubiquitous, especially outside of the North American market. This capped Argmax SDK's maximum attainable device footprint.
• TFLite NPU runtime did not support ahead-of-time (AOT) compilation, requiring apps to run a costly on-device compilation step before the end users could enjoy full app functionality. This severely impacted the first-time user experience and turned out to be a showstopper for many use cases.

Google LiteRT supplanted TFLite as the next-gen inference runtime for Android. Most importantly:

• LiteRT consolidated various chip vendors to support NPU deployment on Qualcomm Snapdragon, Google Tensor and MediaTek chips. These three hardware vendors combined represent the vast majority of devices on the market today with more vendors potentially becoming available in the near future.
• LiteRT brings ahead-of-time (AOT) compilation capabilities to the Android market, allowing tool builders like Argmax to offer pre-compiled models. This addressed the single biggest technical limitation of NPU deployment.

‍

Thanks to our collaboration with Google LiteRT, Argmax became the first commercial SDK to ship with LiteRT that supports virtually all of the major NPU vendors in the Android market.

‍

What is in Argmax Pro SDK for Android?

Since our first foray into Android, Nvidia Parakeet overtook OpenAI Whisper as the leading speech-to-text model. We built the first and only real-time streaming implementation of Nvidia Parakeet on the market, reusing the techniques we published in our ICML 2025 paper. This algorithm elevates the real-time transcription accuracy to match the reference transcription accuracy on pre-recorded audio. Now, the exact same accuracy is available across Android and Apple platforms via Argmax Pro SDK.

‍Argmax Pro SDK for Apple has recently expanded to support real-time transcription with speakers and custom vocabulary. Argmax Pro SDK for Android is expected to reach feature parity over time. The launch version only supports real-time transcription.

Finally, we redesigned Argmax Pro SDK for Android from the ground up to be Kotlin-first for an unapologetically platform-native developer experience with top-tier reliability and familiar abstractions for Android teams.

‍

Try Argmax Pro SDK for Android today:

• Install Argmax Playground from Firebase to test accuracy and performance without writing any code
• Start your 14-day Argmax Pro SDK trial today and prototype it in your app

‍

‍