Bitrate is the amount of data used to represent one second of audio, measured in kilobits per second (kbps). It tracks both quality and file size: a higher bitrate keeps more detail, while a lower bitrate compresses harder and can throw away information a speech recognizer depends on. It is a different quantity from sample rate, though the two are connected.
How bitrate works
For uncompressed audio, bitrate follows a fixed formula: sample rate times bit depth times the number of channels. A 16 kHz, 16-bit, single-channel PCM stream therefore runs at 256 kbps, and there is no choice in the matter once those settings are fixed.
Lossy codecs invert the relationship: you set a target bitrate, and the codec decides which acoustic detail to discard to hit that budget. Opus spans roughly 6 kbps to 510 kbps, while the telephony codec G.711 sits near 64 kbps.
Why bitrate matters for speech-to-text
Low-bitrate compression strips high-frequency detail and adds artifacts, and both push up word error rate. Voice and telephony pipelines often hand over heavily compressed audio, so there is a practical floor below which accuracy starts to slide.
The damage compounds across steps: audio that is compressed, decoded, and recompressed loses a little more each pass, and the recognizer never gets back what earlier stages threw out.
Practical bitrate guidance
Carry the highest bitrate your pipeline can afford, and cut out transcoding steps that stack losses. Codec choice matters as much as the number, because a modern codec preserves speech intelligibility at a given bitrate far better than an older one does.
When you evaluate a provider, test on the actual bitrate your production path produces, not a pristine studio file, or the results will flatter the model and mislead the decision.
Frequently asked questions
Does a higher bitrate mean better quality? Up to a point. Higher bitrate keeps more detail, but past the level a codec needs to encode speech cleanly, extra bits add file size without audible or measurable gains.
What is a good bitrate for speech? For voice, a modern codec around 16 to 24 kbps is usually clean enough for transcription. Telephony's G.711 at 64 kbps is fine; the limiting factor there is the 8 kHz sample rate, not the bitrate.
How does bitrate affect transcription accuracy? Very low bitrates remove high-frequency detail and add compression artifacts, both of which raise word error rate. The effect is worst when audio is compressed more than once along the path.
What does low bitrate mean for audio? It means fewer bits per second and heavier compression, so the file is smaller but coarser. Speech may still be intelligible while losing the fine detail a recognizer uses.