Audio Sample Rate Explained: 44.1 kHz vs 48 kHz (2026)
A CD holds audio at 44.1 kHz. An iPhone writes the sound on a video at 48 kHz. Both numbers describe how often a converter measures the waveform, and the gap between them explains a specific failure that surprises people every week: a clip that plays about eight percent fast and roughly a semitone and a half sharp, because software read the rate number instead of resampling the audio.
The audio sample rate is how many times per second a converter takes a measurement of the waveform. Double the rate keeps double the high frequencies, and nothing beyond that. It is not a quality dial the way bitrate or bit depth are, and treating it as one is where most advice on the subject goes wrong.
This guide covers what 44.1 and 48 kHz actually do, how to check which rate a file uses, how to change it without damaging the sound, and the cases where the correct answer is to leave the file exactly as it is.
The Direct Answer: 48 kHz for Video, 44.1 kHz for Music
If the audio will sit next to picture, use 48 kHz. If it is a music release, use 44.1 kHz. If nothing else will ever open the file, leave it alone.
The reason 48 kHz suits video is arithmetic. It divides evenly into 24, 25, 30, 50 and 60 frames per second: 2,000 samples per frame at 24 fps, 1,920 at 25, and 1,600 at 30. A 44.1 kHz track gives 1,837.5 samples per frame at 24 fps, so the number never lands on a frame boundary, and the editor has to resample to hold sound and picture together.
Music uses 44.1 kHz because the CD defined it and the entire delivery chain still assumes it. Upsampling a finished song to 96 kHz before shipping a 44.1 kHz file adds size and nothing audible.
| Rate | Theoretical ceiling | Where it belongs |
|---|---|---|
| 44.1 kHz | 22.05 kHz | CD and music streaming |
| 48 kHz | 24 kHz | Video, film and broadcast |
| 96 kHz | 48 kHz | Heavy slow motion and sound design |
| 192 kHz | 96 kHz | Ultrasonic capture and archiving |
One clarification that settles most arguments: bit depth is the other control. 24-bit audio gives more headroom above the noise floor. It does not extend the frequency range. A 24-bit 44.1 kHz recording and a 16-bit 96 kHz recording solve different problems, and neither one replaces the other.
What to Have Ready Before You Convert Anything
Three things: a copy of the file, a way to read its current rate, and a decision about the destination. Skipping the third step is how people end up converting 48 kHz down to 44.1 kHz and then back again, losing a little each way.
Convert from the best source you still have. Re-encoding an already-compressed MP3 or AAC adds one more generation of loss on top of the one that already happened. The original camera file, WAV, or project export is the right starting point.
If the audio you need is locked inside a video file, pulling the audio out first is simpler than converting the whole thing. FFmpeg does the work in the commands below. It is free on Windows, macOS and Linux, and several editors bundle it internally, so nothing here requires paid software.
How to Check the Audio Sample Rate of a File
Ask the file what it is before changing anything. This single ffprobe call prints the rate and channel count of the first audio stream:
ffprobe -v error -select_streams a:0 -show_entries stream=sample_rate,channels -of default=nw=1 input.mp4On a 44.1 kHz source the output reads sample_rate=44100, followed by channels=1 or whatever the file actually contains. If the number surprises you, you have usually found the problem.
On the desktop, VLC reports the same details under Tools, then Media Information, on the Codec details tab. Editors show it in the media bin properties. For a folder of clips, loop the same ffprobe call instead of opening files one by one, because the odd clip out is rarely the one you suspect.
Method 1: Change the Audio Sample Rate With FFmpeg
For a standalone audio file, one flag does the work:
ffmpeg -i input.wav -ar 48000 output.wavThe -ar flag sets the output rate, and FFmpeg resamples to reach it. A 16-bit PCM file stays 16-bit; bit depth is untouched unless you ask for a different one, so this command changes the rate and only the rate.
For a video, copy the picture and re-encode only the sound. This tested command leaves the video stream bit-identical and writes 48 kHz audio into the same container:
ffmpeg -i input.mp4 -c:v copy -c:a aac -b:a 192k -ar 48000 output.mp4-c:v copy is the part that matters. Without it FFmpeg decodes and re-encodes the video too, which costs time and a generation of quality for a change that concerns only the audio stream.
FFmpeg’s built-in resampler is adequate for most work. If your build includes the SoX resampler, this variant is measurably cleaner on difficult material such as cymbals and applause, at the cost of speed:
ffmpeg -i input.wav -af aresample=resampler=soxr -ar 48000 -c:a aac -b:a 192k output.m4aCheck whether SoX is compiled in by running ffmpeg -h filter=aresample and looking for soxr in the option list. If it is missing, use the default and stop worrying about it, because the difference is subtle rather than dramatic. FFmpeg’s resampler documentation lists the available settings.
Method 2: Fix the Rate in an Editor, Not the Video
Most editing timelines carry a single audio rate, set once in the sequence or project settings. In Premiere Pro, DaVinci Resolve and Final Cut it sits in those project or sequence audio settings; some editors label it plainly as Sample Rate and others bury it two panels deep.
Set the timeline to 48 kHz, then convert the source clips to 48 kHz on disk before you import them. This is the habit professionals keep, and it exists for a reason: a timeline that resamples on the fly works fine until a long project drifts or a render fails late in a deadline.
The export panel takes the rate from the project, but it can silently override it. Confirm that setting before a 40-minute render rather than after. For a phone-recorded voiceover, record at 48 kHz to match the video instead of accepting whatever a music app defaults to.
What Sample Rate Conversion Does, and What It Cannot Do
Resampling is interpolation. A good converter filters the signal first, then calculates new samples between the existing ones. A poor one repeats or drops samples, which folds unwanted frequencies back into the audible range as aliasing, heard as a gritty edge on sibilance and cymbals.
The hard limit is the sampling theorem: a file can only carry frequencies below half its sample rate. A 44.1 kHz recording holds nothing above 22.05 kHz. Converting it to 48 kHz produces a file that still holds nothing above 22.05 kHz. The extra bandwidth is empty by definition.
That leads to a point most guides skip. The audio sample rate is not what makes a lossy file large: bitrate is. Raising the rate on an AAC or MP3 encode does not raise the bitrate, so the encoder spends most of the new high-frequency space on detail nobody can hear. Size follows bitrate, not rate.
Browsers add another layer. Web Audio routes every sound through one context rate, usually 48 kHz, so a mismatched file gets resampled by the browser before it reaches your ears. The audio context documentation is explicit that playback is resampled to the context rate. On Windows, the same thing happens in shared mode, where the device format decides.
Bluetooth is worse. Most codecs operate at 48 kHz internally, so a 44.1 kHz file is often resampled twice before it reaches wireless earbuds. The loss at each stage is small, but the chain is longer than most people assume.
None of that is a defect, and it is not a reason to convert everything in advance. The practical takeaway is narrower: convert once, at the point where the format matters, and never stack conversions hoping the result improves.
Rights and Responsible Use
Recorded audio usually carries two separate rights, not one: the composition and the recording. A sample rate change touches neither. It is a technical operation performed on a copy, not a licence, and it grants nothing you did not already hold.
Keeping a copy for private listening or for editing your own material is a different act from publishing that audio to an audience. Audio you recorded yourself is yours to convert, publish and license as you choose.
Audio that arrived from somewhere else stays someone else’s work no matter what rate it is stored at, and the terms of the platform you publish to govern what you may upload. YouTube’s Terms of Service set out those rules for the most common destination of a finished video.
If a video needs music you do not own, use the licensed library the platform provides or clear the track with the rights holder. Re-encoding it, trimming it, or matching the sample rate to your project changes none of that.
Troubleshooting: When Resampling Goes Wrong
The audio plays fast and sounds sharp
Software read the rate number instead of resampling the samples. A 44.1 kHz file played as if it were 48 kHz runs about 8.1 percent fast and around 1.5 semitones high, so three seconds of audio finishes in 2.76 seconds. The repair is a real conversion with -ar, not an edit to a metadata field.
Sound drifts further out of sync as the file plays
That is clock drift, not a rate mismatch. A wrong rate produces a constant offset that never grows; drift that accumulates points at a different cause, usually a variable frame rate recording. Separating those two faults is the first step of the repair.
Crackle or a metallic edge appears after conversion
Two likely causes. A resampler without proper anti-aliasing smears high frequencies downward, and interpolation can push peaks slightly past full scale, so a track already sitting at 0 dBFS distorts once it is converted. Convert into a floating-point format and level the result afterwards, which costs nothing and prevents the worst of it.
The converted file plays differently on a phone or a TV
The player is resampling to its own hardware rate, or the container advertises a rate the decoder misreads. That is a playback chain issue rather than file corruption. Re-mux the file at a common rate, then test in a player that reports stream details so you can see what the device actually receives.
A second conversion made the top end worse
Generation loss. Each lossy pass discards detail permanently, and converting an MP3 twice compounds it. Go back to the original recording and convert once from there, even if that means exporting a fresh mix first.
The export has no audio, or the wrong track
Usually a stream mapping mistake rather than a rate problem. A file with several audio streams can end up with the wrong one mapped after adding -ar. Name the stream explicitly with -map 0:a:0 to pin the first audio track to the output.
Frequently Asked Questions
Does a higher audio sample rate sound better?
No, not on ordinary playback. Once the rate clears the highest frequency a recording contains, the extra samples carry nothing audible. A 96 kHz copy of a 44.1 kHz master sounds the same as the original.
Is 48 kHz better than 44.1 kHz?
For video, yes. For music delivery, no. The right rate depends on what will consume the file, not on which number happens to be larger.
What sample rate should I record at?
48 kHz for anything that will sit beside picture, voiceovers included. Use 44.1 kHz only when the recording is music heading for a music release. Recording above 48 kHz is a deliberate choice for material you plan to slow down later, not a general upgrade.
Does changing the sample rate make a file smaller?
Not for compressed formats. AAC and MP3 size follows the bitrate you set, so 48 kHz and 44.1 kHz encodes at 192 kbps land at nearly the same size. Uncompressed PCM is the exception: dropping from 48 kHz to 44.1 kHz removes about 8 percent of the audio data.
Can I convert 44.1 kHz to 48 kHz without losing quality?
You keep everything the file already contained. What you cannot do is create detail that was never captured, because those frequencies are absent from the recording. Converting from the original source is the only way to recover them.
Why does my audio sound high-pitched after converting?
The rate was relabelled rather than resampled. That shortcut speeds playback by roughly 8 percent and raises pitch by about 1.5 semitones, which is immediately obvious on a human voice.
Do I need 96 kHz or 192 kHz?
Rarely. They earn their place when you intend to slow audio down heavily or shift pitch by large intervals, because the extra bandwidth survives the change. For ordinary playback they add file size and processing load for nothing audible.
Which Sample Rate Should You Use?
The choice comes down to where the file is going. Audio that will be edited against picture wants 48 kHz, music heading for release wants 44.1 kHz, and material you intend to slow down dramatically justifies 96 kHz.
If none of those apply, do nothing. A file at the wrong rate for a given player is resampled by that player, usually without complaint. Converting is worth the time only when the mismatch causes a problem you can actually hear or see.
So: match the rate to its destination, start from the best source you still have, and convert once.