Measure and compare in three steps
Choose an actual exported file and keep its complete programme intact. No account is needed. The first run loads the analysis engine, then decodes the selected audio stream in a worker within your browser.
- 1
Choose audio or a video soundtrack
Select one local file and press Measure loudness. The first audio stream is analyzed from beginning to end. A video without an audio stream produces a clear error instead of an empty success report. Cancel stops the job; reset lets you choose another source.
- 2
Read the summary and timeline
Integrated loudness describes the gated programme level in LUFS. Loudness range describes variation in LU, while true peak estimates reconstructed peak level in dBTP. The graph separates 400 ms momentary readings from three-second short-term readings rather than presenting a single moving volume bar.
- 3
Compare and save the evidence
Select a comparison target and peak ceiling. The resulting gain estimates update without decoding the file again. Download JSON for the summary and comparison or CSV for every 100 ms timeline interval. Copy JSON when you need to share results in a review or issue.
Why peak level cannot replace loudness
A single transient can set a high peak even when most of a programme is quiet. Conversely, a dense mix may sound much louder without having a higher maximum peak. Integrated loudness and true peak answer different questions, so the report shows both instead of suggesting that one percentage describes perceived level.
Loudness range adds context about variation across the programme. A low range is not automatically better, and a high range is not automatically a fault. The appropriate result depends on the material and listening conditions. Use the timeline to find the section that explains an unexpected reading before applying processing to the entire file.
Comparison targets and peak headroom
The available targets are -14, -16 and -23 LUFS, with -1 or -2 dBTP ceilings. These are numerical comparison choices, not promises about any service's current delivery rules. Gain to target equals target minus measured integrated loudness. Gain within the peak ceiling is the smaller of that requested gain and the available peak headroom. A negative result means attenuation, not extra gain. Missing gated loudness or unavailable peak information remains unknown rather than being replaced with zero.
Useful before your final audio delivery
Read short files and silence carefully
The engine emits timeline metadata at 100 ms intervals. The 400 ms and three-second windows need time to fill, so their startup intervals are blank. Files shorter than three seconds do not provide representative short-term or loudness-range information. Silence and signals below the integrated gate can produce no meaningful integrated result. A dash means unavailable or not measurable, never a silent file at 0 LUFS.
- Measure the complete intended programme. Selecting only an introduction or chorus can produce a very different integrated result from analyzing the entire delivered recording.
- Keep loudness and peak headroom separate. A quiet mix can still contain a sharp transient close to the ceiling, leaving little room for uniform gain without further processing.
- Re-measure the final encoded delivery rather than assuming it matches a source master. Lossy encoding and later edits can affect peaks or change the balance of loud and quiet sections.
Your recording stays in the tab
Analysis runs locally using FFmpeg in a browser worker. The engine files are downloaded when required, but your selected recording is not uploaded by this tool. The optional player uses a local object URL and depends on the browser's playback codecs. A preview error does not necessarily prevent the analysis engine from reading the file.
Downloaded reports contain measurements, timeline data and the current comparison, not the audio itself or the original filename. Sharing a report is still your decision. Reset removes the current selection and result from the interface, while the original recording on your device stays unchanged.
How to use a loudness report
A spoken episode measures below your chosen target while its true peak is already near the ceiling. The requested gain and peak-limited gain differ, showing why a simple volume boost may not be sufficient.
An interview starts quietly and introduces music halfway through. The momentary trace reacts quickly, while the short-term trace shows a smoother change over the surrounding three seconds.
A video contains several languages. This page measures the first audio stream only; inspect the track order first and use a separate file for each additional mix that needs its own report.
Measurement boundaries
Only the first audio stream is measured; there is no automatic language or commentary-track selection. Protected files, remote URLs and playlists are outside scope. Very large recordings may exceed browser memory. Summary measurements are rounded to one decimal, the chart may downsample long timelines, and the CSV retains all emitted intervals. An incomplete final interval is not displayed. Channel weighting follows the engine; this tool does not apply mono-to-dual-mono compensation. Recheck critical delivery work with the required reference meter.
- Do not treat this as a live studio meter, hardware calibration instrument or broadcast compliance certificate. It analyzes a local file after recording, using the browser engine.
- Use an audio processor when you need limiting, compression or normalization. This page calculates measurements and a gain estimate, but does not create a louder or quieter audio file.
Loudness measurement questions
Will choosing -16 LUFS normalize my file?
No. The target changes only the comparison and estimated gain. No audio is rendered, limited or altered. Use an audio normalization tool to create a processed version, then measure that exported version again to verify the result.
Why can the peak-limited gain be smaller than the target gain?
Uniform gain raises peaks along with the programme level. If a transient is already close to the chosen ceiling, reaching the loudness target would push it above that ceiling. The estimate reports this constraint; it does not predict what a limiter or compressor would do.
Why is the beginning of the graph blank?
Momentary and short-term measurements use different rolling windows. There is not enough preceding audio at startup to fill them, especially the three-second window. These gaps are kept blank instead of drawing the engine's startup sentinel as a real loudness value.
Are these targets official settings for streaming platforms?
No. Delivery requirements vary and can change, and different destinations may use different playback behavior. Treat the targets as comparison numbers you deliberately choose. Check the receiving service's current specification before a professional delivery.