The online encoder uses a standalone LAME 3.100 WebAssembly bridge. This page answers a narrow question: does it behave like the command-line encoder family, or does the browser version introduce obvious timing, frequency-response or quality problems?
The browser encoder looks consistent with normal LAME behaviour. The tests did not show clock drift, accidental 48 kHz resampling, or a browser-specific quality collapse.
VBR V0 behaves like a high-quality modern VBR encode, VBR V2 behaves like an efficient practical default, and the 192 kbps modes look like the expected bitrate-constrained command-line family.
The browser version is convenient for single-file work. The command-line builds remain better for batch encoding, scripting, exact command-line switches, cVBR/cVBRb builds, and experimental branch features.
More negative Difference RMS values and higher correlated-null values are better in this context. These are practical comparison figures from a small test set, not a formal codec listening test.
| Browser mode | Difference RMS | Correlated null | Frequency-response read | Measurement summary |
|---|---|---|---|---|
| CBR 320 | -48.34 dB | 58.13 dB | Stronger roll-off above roughly 19-20 kHz than VBR V0. | Lowest Difference RMS and deepest correlated null in this test set. |
| VBR V0 | -45.90 dB | 55.70 dB | Strongest upper-treble retention in the web tests; useful energy remains into about 20 kHz. | Closest match to the stronger modern command-line VBR behaviour. |
| CBR 192 | -39.73 dB | 48.23 dB | Broadly similar to the 192 kbps command-line family; upper treble begins falling around the 16 kHz region. | Sensible 192 kbps result; slightly ahead of ABR 192 in this sample. |
| ABR 192 | -38.93 dB | 47.89 dB | Similar 16 kHz-region roll-off to command-line ABR/192 kbps tests. | No anomalous timing or frequency-response behaviour observed. |
| VBR V2 | -38.73 dB | 48.42 dB | Typical modern LAME V2-style tradeoff: efficient, but upper-treble retention below V0. | Lower-bitrate VBR option with reduced upper-treble retention compared with V0. |
| CBR 128 | -33.34 dB | 40.95 dB | Clearly the most bitrate-limited result. | Expected low-bitrate compromise. |
The upper-band results line up with the earlier command-line lab work: higher-quality VBR preserves more upper treble, while the 192 kbps and lower modes apply a stronger practical low-pass tradeoff.
| Browser mode | Mean loss 16-18 kHz | Mean loss 18-20 kHz | Mean loss 20-21.5 kHz | 3 dB drop | 6 dB drop |
|---|---|---|---|---|---|
| VBR V0 | -0.8 dB | -2.4 dB | -7.8 dB | ≈19.29 kHz | ≈20.16 kHz |
| CBR 320 | -2.0 dB | -4.4 dB | -35.8 dB | ≈17.37 kHz | ≈19.44 kHz |
| VBR V2 | -5.8 dB | -29.6 dB | -38.1 dB | ≈16.01 kHz | ≈16.07 kHz |
| CBR 192 | -11.7 dB | -33.8 dB | -38.9 dB | ≈16.01 kHz | ≈16.03 kHz |
| ABR 192 | -18.0 dB | -36.0 dB | -38.7 dB | ≈15.99 kHz | ≈16.01 kHz |
| CBR 128 | -41.0 dB | -43.7 dB | -38.6 dB | ≈15.42 kHz | ≈15.99 kHz |
VBR V0 was the strongest browser result for upper-treble retention. CBR 320 had the best residual/null result overall, but the parsed spectrum showed stronger roll-off above roughly 19-20 kHz. That is a useful reminder that “best null” and “widest top end” are related, but not identical, measurements.
These results describe the LAME 3.100 WebAssembly encoder build tested for this comparison. They are based on a limited source set and are intended to identify major differences in timing, frequency response and encoded output, rather than provide a formal listening-test ranking.
The page does not currently record a WebAssembly bridge revision, package hash or test date. The measurements should therefore be repeated whenever the browser encoder, WebAssembly bridge or underlying LAME build changes.
For a quick one-off encode, use the browser encoder. For a practical default, VBR V2 is a good starting point. For a higher-bitrate browser encode, use VBR V0. For maximum compatibility or fixed-rate delivery, use CBR. For serious batch work or experimental branch testing, use the downloadable command-line builds.