Sample Rate Converter

Resample audio to 8, 16, 44.1 or 48 kHz โ€” the speech-to-text prep tool.

About this converter

Pick a target rate and the browser's resampler does the maths with proper filtering โ€” no aliasing artifacts. The status line tells you honestly what each direction means: what downsampling discards, and what upsampling doesn't restore.

Nyquist, and why 16 kHz is the speech magic number

A sample rate captures frequencies up to exactly HALF itself โ€” the Nyquist limit, and the single most useful fact in digital audio. From it, every standard rate explains itself: 44.1 kHz covers hearing's ~20 kHz ceiling (hence CDs), 48 kHz is video's standard with margin, 16 kHz keeps everything below 8 kHz โ€” which is where nearly all speech energy lives, which is why 16 kHz mono is THE format transcription APIs and speech models ask for (mono conversion is the other half of that prep). 8 kHz is telephone quality: speech survives, sibilance dulls โ€” the "phone voice" sound is literally this number. Two honest one-liners the status repeats: downsampling genuinely discards the frequencies above the new Nyquist (by mathematics, not carelessness), and UPsampling never restores them โ€” it changes the container, not the content, useful only for format-matching. Bonus: WAV size scales with rate, so 44.1 โ†’ 16 kHz cuts nearly two-thirds โ€” the size calculator does the arithmetic for any duration.

Frequently asked questions

What rate do transcription services want?

16 kHz mono is the overwhelming standard for speech-to-text โ€” this tool plus stereo-to-mono produces exactly that. Speech loses nothing that matters; files shrink by two-thirds.

Does upsampling improve audio?

No โ€” it changes the container, not the content. Frequencies the original rate never captured cannot be computed back. Upsample only to match a format requirement.

What actually disappears when I downsample?

Everything above half the new rate (Nyquist): at 16 kHz, all content above 8 kHz โ€” cymbal sheen and sibilant sparkle, but essentially no speech information.

Why do CDs use 44.1 kHz specifically?

Half of it (22.05 kHz) clears human hearing's ~20 kHz ceiling with engineering margin; the odd number itself is a relic of storing early digital audio on video equipment. 48 kHz won video for tidier maths.