How much data does an AI voice recorder produce and how fast does it transfer
A wearable AI recorder using 16 kbps Opus encoding produces about 7.2 MB of data per 1 hour. On modern smartphones, transferring a 1-hour recording takes roughly 4 minutes via Bluetooth Low Energy. This article details the exact measurements, storage calculations, and platform-specific transfer speeds for product managers sizing storage and evaluating integration.
What is an AI voice recorder data profile
An AI voice recorder data profile is the exact combination of audio encoding bitrate, resulting file size per unit of time, and the wireless throughput required to move that file to a host device. For our AI voice recorder hardware, this profile is defined by 16 kbps Opus frames, yielding 7.2 MB per 1 hour, transferred over Bluetooth Low Energy at approximately 30 KB/s on mobile platforms. Understanding this profile is the first step in sizing local storage and planning your application architecture.
How we measured the data size and transfer speed
We tested a CB08 hardware unit running firmware version 1.0.0. The device was fully charged and recorded continuously in a standard office environment. We measured the exact byte count of the resulting files and timed the transfer to 3 different host platforms. We did not test Wi-Fi transfer, as this device relies exclusively on Bluetooth Low Energy. You can read more about how an AI note taker device works in our technical overview.
The test hosts were:
- Android 9 smartphone (Xiaomi MI 6) using a native test application with high-priority connection parameters.
- iPhone 16 using a third-party application.
- Windows 10 PC using both a Chrome web browser (Web Bluetooth) and a native Python script.
Every file size and transfer time in this article comes directly from these test logs. We recorded a continuous 60-minute and 1-second audio file. The device stored exactly 7202040 bytes for this file.
Why Opus at 16 kbps for voice recording
The device records audio using the Opus codec, as defined in RFC 6716. We chose a fixed bitrate of 16 kbps because it provides excellent speech intelligibility while minimizing storage and transfer overhead. The device stores raw Opus frames, not a container format like Ogg or MP3.
Each frame is exactly 40 bytes and represents 20 milliseconds of audio. The math is straightforward: 40 bytes multiplied by 50 frames per second, multiplied by 8 bits per byte, equals 16000 bits per second. This results in a highly predictable file size. We evaluated uncompressed WAV format (PCM, 16 kHz, 16-bit, mono), but it produces about 115 MB per 1 hour. We do not use WAV for storage because it wastes space and makes Bluetooth transfer impractical.
Storage math for 1 hour, 8 hours, and 100 hours
Product managers often need to know how much local storage is required for continuous recording. The device offers 32 GB and 64 GB storage options. Because the Opus bitrate is fixed, we can calculate the exact storage requirements for any duration.
| Recording Duration | Data Size (Calculated) | Notes |
|---|---|---|
| 1 hour | 7.2 MB | Measured directly as 7202040 bytes. |
| 8 hours | 57.6 MB | Calculated as 7.2 MB multiplied by 8. |
| 100 hours | 720 MB | Calculated as 7.2 MB multiplied by 100. |
Even at 100 hours of continuous recording, the data footprint is 720 MB. A 32 GB card can hold about 4444 hours of audio. This means storage capacity is rarely the bottleneck in product design; battery life and transfer speed are the primary constraints.
Bluetooth transfer results by platform
Transfer speed depends entirely on the host device and its Bluetooth Low Energy connection parameters. The recorder itself is the bottleneck, capable of sending a maximum of about 200 packets per second, where each packet is 160 bytes. The host must negotiate the correct MTU and connection interval to achieve optimal speeds.
| Host Platform | Average Speed | Time to Transfer 1 Hour (7.2 MB) |
|---|---|---|
| Android (High Priority) | 30.7 KB/s | 3 minutes 49 seconds |
| iPhone 16 | 28 to 33 KB/s | About 4 minutes |
| Windows 10 (Python) | 2 to 18 KB/s | About 22 minutes (with restart) |
On Android, setting the MTU to 247 bytes and requesting a high-priority connection interval allows the transfer to complete in under 4 minutes. The iPhone performs similarly. The transfer protocol supports segmented downloads and resume from a specific byte offset, which is critical for handling connection drops on mobile devices.
Why Windows was slower and unstable
When testing on a Windows 10 PC, the transfer speed fluctuated between 2 KB/s and 18 KB/s. Transferring the full 1-hour file took approximately 22 minutes and required 1 manual restart of the transfer process. The speed dropped significantly during the first 36 percent of the transfer, averaging only 2.7 KB/s, before recovering to 13.6 KB/s after a reconnect.
We state these observations as measured facts. The exact cause of the instability on Windows 10 was not isolated during this test cycle. Windows does not allow user-space applications to modify Bluetooth connection intervals, which limits our ability to optimize the link layer from the host side. For product integrations, we recommend targeting mobile platforms for primary file transfer, or ensuring your Windows application implements robust retry and resume logic.
What this means for product design and integration
The data profile of this recorder directly informs several hardware and software decisions. First, the 400 mAh battery provides 26 to 30 hours of continuous recording. Since the data size is only 7.2 MB per 1 hour, the device can easily record for its entire battery life without filling the 32 GB storage. Second, the reliance on Bluetooth Low Energy means your application must handle connection drops gracefully. Implementing the segmented download command and offset-based resume is mandatory for a stable user experience.
Third, the raw Opus format requires your application to encapsulate the frames into an Ogg container before passing them to standard media players or speech-to-text APIs. Finally, because the transfer speed on mobile is roughly 30 KB/s, a user returning from an 8-hour shift will wait about 32 minutes for the audio to sync to their phone. Your application design should handle this transfer in the background and provide clear progress indicators.
FAQ
How much storage does 1 hour of recording use?
1 hour of recording uses exactly 7.2 MB of storage when encoded at 16 kbps Opus. This was measured directly from the device file system.
How long does it take to transfer a 1-hour recording to a phone?
On modern Android and iPhone devices, transferring a 1-hour recording takes about 4 minutes. This is based on measured transfer speeds of approximately 30 KB/s over Bluetooth Low Energy.
Why is the transfer speed slower on Windows computers?
Measured transfer speeds on Windows 10 fluctuate between 2 KB/s and 18 KB/s, taking about 22 minutes for a 1-hour file. Windows restricts applications from modifying Bluetooth connection parameters, which limits throughput optimization.
Does the device record in WAV or MP3 format?
The device records in raw Opus frames. It does not use WAV or MP3. The raw Opus frames must be encapsulated into an Ogg container by your application before playback or transcription.
Can the device transfer data while it is recording?
Transferring data while simultaneously recording has not been tested. We do not make commitments on this behavior until it has been verified in a controlled test environment.
About this page
This page is published by PINGOOD TECHNOLOGY CO. LIMITED. The scope of this article covers the data size, storage requirements, and Bluetooth transfer speeds of our CB08 AI voice recorder hardware. The file sizes and transfer times for Android, iPhone, and Windows 10 are measured data taken from our internal test logs. The storage requirements for 8 hours and 100 hours are calculated data derived by multiplying the measured 1-hour file size by the respective number of hours. Hardware specifications such as battery capacity and storage options are based on our official product facts.
Related
Want to try the hardware behind these notes?
Six AI hardware platforms are in mass production. Tell us your market and channel and we send an evaluation sample.
Request a sample