Realtime adjuster results
LevelAPI2 202609280, Medium preset, measured by the published test method on 113 cases at -23 LUFS and 67 programme cases at -16 LUFS: synthetic level jumps and pauses, broadcast recordings, films, music, read speech, conversation and a live stream: 25 hours of material. Every graded line is shown with every case that fails it and the measured figure.
| Declaration | Value |
|---|---|
| Device | LevelAPI2 realtime adjuster, revision 202609280, preset Medium, every other setting at its default |
| Target | -23 LUFS; the programme cases again at -16 LUFS |
| True-peak ceiling | -1 dBTP, the default (SetTruePeakCeilingDb) |
| Latency | 50 ms: 2400 frames at 48 kHz, 2205 at 44.1 kHz, measured on the audio |
| Maximum boost | +12 dB |
| Reference block | 10 ms (480 frames at 48 kHz, 441 at 44.1 kHz) |
| Block sizes declared identical | every size. Tested at 20 ms and 256, 512, 1024 and 2048 frames on every case but P6 and the conversation cases, and at 64 frames on the synthetic cases and the EBU / ITU files |
| Program length | the whole stream by default; the long cases again with 900 s |
The five pass lines of the method, section by section. A failing row stays on this page with its figure and what it is.
| Pass line | Scope | Result |
|---|---|---|
| G1 | Whole-case integrated output within ±1 LU of ‑23 LUFS, every programme case | 68 of 68, and 15 of 15 with a 900 s program length; all pass at every other block size |
| G2 | Long-term loudness in band on every case longer than 16 min | 13 of 15; 14 of 15 with a 900 s program length |
| G3 | Maximum true peak at or under ‑0.8 dBTP on the lossless cases (synthetic signals and the EBU / ITU files), at every block size | 47 of 49 at every block size; the highest passing case reads ‑0.99 dBTP |
| G4 | The last 30 s of each constant synthetic part within ±1 LU, where the need is within the +12 dB maximum | 67 of 76, the same parts at every block size |
| G5 | The output at every other block size identical to 10 ms blocks, sample for sample | 20 ms 108 of 108; 64 frames 49 of 49; 256 frames 108 of 108; 512 frames 108 of 108; 1024 frames 108 of 108; 2048 frames 107 of 108 |
| Line | Case | Material | Measured | What it is |
|---|---|---|---|---|
| G2 | P12 | Live stream 1, game audio and voice | 87.03 % in band | A sparse live stream: over its quiet stretches (-35 to -58 LUFS) the gain sits on the +12 dB maximum and the output stays low. The fixed-gain two-pass reference reads 69.1 % here. |
| G2 | P6 | Long run, four mixes back to back (workout, mixed, pop, rock) | 92.24 % in band | Three hours and 42 minutes of music with no program length: after two hours a new minute is too small a share of the long-term correction to move it. With 900 s it passes. |
| G3 | S6.1 | pink T-15, 10 s silence, bursts at T | 10 ms ‑0.46, 20 ms ‑0.46, 64 frames ‑0.46, 256 frames ‑0.46, 512 frames ‑0.46, 1024 frames ‑0.46, 2048 frames ‑0.46 dBTP | White-noise bursts after a quiet passage and digital silence. The limiter holds -1 dBTP on its own 4x estimate; a BS.1770 meter reads these bursts up to about 0.6 dB higher. The ceiling setting is there for this: set it that much lower for a guarantee on such material. |
| G3 | S6.2 | pink T-15, 10 s room tone at -50, bursts at T | 10 ms ‑0.46, 20 ms ‑0.46, 64 frames ‑0.46, 256 frames ‑0.46, 512 frames ‑0.46, 1024 frames ‑0.46, 2048 frames ‑0.46 dBTP | As S6.1, with room tone at -50 LUFS instead of silence. |
| G4 | S3.up part 2 | P5, -12 dB -> 0 dB at 180 s | ‑24.89 LUFS | Programme material split at its file joins, 30 to 83 s after a 12 dB step: the ride has not settled yet. |
| G4 | S3.up part 3 | P5, -12 dB -> 0 dB at 180 s | ‑20.50 LUFS (the fixed-gain reference reads ‑20.50) | Programme material split at its file joins, 30 to 83 s after a 12 dB step: the ride has not settled yet. |
| G4 | S3.up part 4 | P5, -12 dB -> 0 dB at 180 s | ‑24.80 LUFS | Programme material split at its file joins, 30 to 83 s after a 12 dB step: the ride has not settled yet. |
| G4 | S3.down part 1 | P5, 0 dB -> -12 dB at 180 s | ‑21.25 LUFS | As S3.up, after a 12 dB step down. |
| G4 | S3.down part 2 | P5, 0 dB -> -12 dB at 180 s | ‑25.20 LUFS | As S3.up, after a 12 dB step down. |
| G4 | S5.-60.30 part 3 | pink T-10, 30 s room tone at -60, pink T-10 | ‑24.13 LUFS | Room tone in a pause of 30 s or more is lifted up to 12 dB. The returning programme meets that boost and is over-corrected afterwards, so the part ends up to 1.6 LU low. |
| G4 | S5.-60.120 part 3 | pink T-10, 120 s room tone at -60, pink T-10 | ‑24.56 LUFS | Room tone in a pause of 30 s or more is lifted up to 12 dB. The returning programme meets that boost and is over-corrected afterwards, so the part ends up to 1.6 LU low. |
| G4 | S5.-50.120 part 3 | pink T-10, 120 s room tone at -50, pink T-10 | ‑24.26 LUFS | Room tone in a pause of 30 s or more is lifted up to 12 dB. The returning programme meets that boost and is over-corrected afterwards, so the part ends up to 1.6 LU low. |
| G4 | S5.-45.120 part 3 | pink T-10, 120 s room tone at -45, pink T-10 | ‑24.07 LUFS | Room tone in a pause of 30 s or more is lifted up to 12 dB. The returning programme meets that boost and is over-corrected afterwards, so the part ends up to 1.6 LU low. |
| G5 | S7 at 2048 frames | pink T-10, reset at 120 s | 85 of 3000 ticks differ from 120.0 s | The test harness resets the stream on the first host block boundary at or after 120 s. At 2048 frames that boundary falls 1024 frames later than at every other size, so the stream is reset at a different sample. Identical again from 140.8 s. |
Where the adjuster is reliable, and what it does outside that range. "What changes it" names only what the shipped API offers, and what a host controls anyway.
| What the listener hears | Case that shows it | What changes it |
|---|---|---|
| After a jump to louder material (an ad after a quiet programme) it is too loud for about half a minute, by up to 7.7 LU, then settles | S1 and S2 steps up: back in band in 26-48 s (S2.20, which arrives on the +12 dB maximum: 85 s) | Not adjustable. A host that knows the boundary (a playlist change, an ad insertion) can call Reset() there: the stream starts fresh from 0 dB (S7: back in band 9.8 s after a reset). Reset() allocates, so call it off the audio thread |
| After a drop to quieter material it stays quiet for up to a minute and a half | S1 and S2 steps down: back in band in 47-83 s | Not adjustable; Reset() at a known boundary as above |
| Material more than 12 LU under the target stays under it | S1.-20 (never back in band), P12's quiet stretches | Not adjustable: +12 dB is the most the adjuster adds |
| At a raised target, quiet material stays under it: at -16 LUFS, anything quieter than about -28 LUFS | -16 LUFS run: P7.02 (‑17.65 LUFS), P13.3 (‑18.79 LUFS) | SetTargetLUFS(): a lower target leaves more material within +12 dB. The same cases pass at -23 LUFS |
| Background noise in a pause is raised (room tone, hiss, an audience), by up to 12 dB when the pause lasts 30 s or more. The programme then returns about 2 LU loud for 25-51 s, then up to 1.6 LU quiet | S5, room tone for 30 s or 120 s. Five seconds of room tone costs nothing | Not adjustable |
| After digital silence: pauses of 10 s or less return unchanged. After 30 s, material that needed a lift returns up to 4 LU quiet for about 6 s; after a minute or more the correction has faded back to 0 dB, and quiet material returns up to 10 LU under the target, back in band in about 12 s | S4, silence of 2 to 120 s. Material at the target returns clean after every length | Not adjustable. A long pause is treated as the end of the programme |
| A stream switched on before its programme starts raises the room tone by 12 dB; the programme arrives up to 7.4 LU loud, then dips, back in band in 24-141 s | S8, room tone for 30 to 300 s before the programme | Start the stream, or call Reset(), when the programme starts |
| A percussive entry after a quiet passage is up to 3.8 LU loud for about 18 s | S6.1, S6.2 | Not adjustable |
| Peaks read above the ceiling on a BS.1770 meter: up to about 0.6 dB on noise-like material, up to about 0.35 dB on lossy programme at -16 LUFS | S6: ‑0.46 dBTP at the default -1 dBTP. At -16 LUFS: P13.4, ‑0.66 dBTP | SetTruePeakCeilingDb(): set the ceiling that much lower. Lossless programme material stays under -0.8 dBTP at the default |
| On a feed that runs for hours, the level drifts up to about 1.3 LU after the first two hours | P6: long-term loudness in band 92.2 % | SetProgramLengthSeconds(900): 99.9 % on the same feed; neutral on an hour of material |
| Peaks above the ceiling with the limiter off | not measured | SetLimiterEnabled(true), the default |
Median share of the time the output's loudness sits within ±1 LU of the target, over 10 s, 30 s and 300 s windows, for the adjuster / a two-pass normalisation that knows the whole file in advance / the untouched input. The two-pass figure is what a file-based workflow achieves; a live adjuster cannot know the future, and trades that for following the programme as it changes.
| Material | Cases | W10 in band % | W30 in band % | W300 in band % | Gain movement dB/min |
|---|---|---|---|---|---|
| Broadcast recordings | 11 | 77 / 66 / 2 | 98 / 77 / 0 | 100 / 100 / 0 | 4.0 |
| Zapping stream | 1 | 62 / 25 / 29 | 79 / 24 / 31 | 98 / 30 / 22 | 3.6 |
| EBU / ITU reference programme | 4 | 24 / 11 / 13 | 29 / 32 / 32 | 100 / 60 / 59 | 15.1 |
| Films | 17 | 23 / 19 / 3 | 43 / 25 / 3 | 94 / 63 / 0 | 17.0 |
| Music mixes | 5 | 58 / 53 / 0 | 69 / 62 / 0 | 95 / 94 / 0 | 2.5 |
| Read speech, long | 3 | 60 / 51 / 3 | 83 / 62 / 1 | 100 / 91 / 0 | 5.4 |
| Poetry readings, short | 20 | 60 / 62 / 0 | 98 / 99 / 0 | 100 / 100 / 0 | 5.6 |
| Conversation | 4 | 58 / 54 / 1 | 81 / 65 / 0 | 100 / 77 / 0 | 5.0 |
| Monologue over a bed | 1 | 55 / 37 / 3 | 72 / 43 / 3 | 99 / 57 / 1 | 4.4 |
| Live stream | 1 | 23 / 22 / 19 | 39 / 24 / 24 | 91 / 69 / 31 | 20.6 |
| Long run, 3 h 42 min | 1 | 56 / 31 / 1 | 65 / 35 / 1 | 85 / 37 / 0 | 2.8 |
Pink noise at defined levels, programme material stepped by 12 dB, silence, room tone and noise bursts. "Need" is the gain a perfect fixed gain would apply after the change; "over" and "under" are the largest excess of the short-term loudness over the minute after it.
| Case | Signal | Change at s | Need dB | Over LU | Under LU | Above band s | Back in band s |
|---|---|---|---|---|---|---|---|
| S1.+6 | pink T -> T+6 | 180 | ‑6.0 | 6.0 | 0.0 | 27.0 | 35.2 |
| S1.+12 | pink T-2 -> T+10 (re-based) | 180 | ‑10.0 | 7.3 | 0.1 | 42.1 | 42.6 |
| S1.+20 | pink T-10 -> T+10 (re-based) | 180 | ‑10.0 | 7.7 | 0.5 | 44.9 | 47.9 |
| S1.-6 | pink T -> T-6 | 180 | +6.0 | 0.0 | 5.4 | 0.0 | 50.2 |
| S1.-12 | pink T -> T-12 | 180 | +12.0 | 0.0 | 11.8 | 0.0 | 71.1 |
| S1.-20 | pink T -> T-20 | 180 | +20.0 | 0.0 | 20.0 | 0.0 | not in the case |
| S2.6 | pink T-6 -> T | 180 | +0.0 | 5.5 | 0.4 | 24.8 | 26.2 |
| S2.12 | pink T-12 -> T | 180 | +0.0 | 7.1 | 0.7 | 38.4 | 39.2 |
| S2.20 | pink T-20 -> T | 180 | +0.0 | 7.2 | 7.1 | 18.2 | 84.9 |
| S2.-6 | pink T+6 -> T | 180 | +0.0 | 0.0 | 5.3 | 0.0 | 47.0 |
| S2.-12 | pink T+10 -> T-2 (re-based) | 180 | +2.0 | 0.4 | 11.4 | 0.0 | 68.3 |
| S2.-20 | pink T+10 -> T-10 (re-based) | 180 | +10.0 | 0.4 | 20.4 | 0.0 | 83.4 |
| S3.up | P5, -12 dB -> 0 dB at 180 s | 180 | ‑1.6 | 4.8 | 2.5 | 32.3 | not in the case |
| S3.up | P5, -12 dB -> 0 dB at 180 s | 246 | +0.0 | 3.9 | 2.2 | 3.7 | not in the case |
| S3.up | P5, -12 dB -> 0 dB at 180 s | 330 | +2.9 | 0.0 | 4.8 | 0.0 | not in the case |
| S3.down | P5, 0 dB -> -12 dB at 180 s | 180 | +10.4 | 1.7 | 8.3 | 4.5 | not in the case |
| S3.down | P5, 0 dB -> -12 dB at 180 s | 246 | +12.0 | 0.0 | 1.1 | 0.4 | not in the case |
| S3.down | P5, 0 dB -> -12 dB at 180 s | 330 | +14.8 | 0.0 | 4.8 | 1.0 | not in the case |
| S4.T.2 | pink T, 2 s silence, pink T | 122 | +0.0 | 0.0 | 0.0 | 0.0 | 0.1 |
| S4.T.10 | pink T, 10 s silence, pink T | 130 | +0.0 | 0.0 | 0.0 | 0.0 | 0.8 |
| S4.T.30 | pink T, 30 s silence, pink T | 150 | +0.0 | 0.0 | 0.0 | 0.0 | 0.8 |
| S4.T.60 | pink T, 60 s silence, pink T | 180 | +0.0 | 0.0 | 0.0 | 0.0 | 0.8 |
| S4.T.120 | pink T, 120 s silence, pink T | 240 | +0.0 | 0.5 | 0.0 | 0.0 | 0.8 |
| S4.L.2 | pink T-10, 2 s silence, pink T-10 | 122 | +10.0 | 0.0 | 0.5 | 0.0 | 0.1 |
| S4.L.10 | pink T-10, 10 s silence, pink T-10 | 130 | +10.0 | 0.0 | 0.5 | 0.0 | 0.8 |
| S4.L.30 | pink T-10, 30 s silence, pink T-10 | 150 | +10.0 | 0.5 | 4.0 | 0.0 | 6.3 |
| S4.L.60 | pink T-10, 60 s silence, pink T-10 | 180 | +10.0 | 1.0 | 10.0 | 0.0 | 11.9 |
| S4.L.120 | pink T-10, 120 s silence, pink T-10 | 240 | +10.0 | 1.0 | 10.0 | 0.0 | 11.9 |
| S5.-60.5 | pink T-10, 5 s room tone at -60, pink T-10 | 125 | +10.0 | 0.0 | 0.5 | 0.0 | 0.1 |
| S5.-60.30 | pink T-10, 30 s room tone at -60, pink T-10 | 150 | +10.0 | 2.0 | 0.1 | 27.7 | 11.1 |
| S5.-60.120 | pink T-10, 120 s room tone at -60, pink T-10 | 240 | +10.0 | 2.0 | 1.0 | 50.8 | not in the case |
| S5.-50.5 | pink T-10, 5 s room tone at -50, pink T-10 | 125 | +10.0 | 0.0 | 0.5 | 0.0 | 0.1 |
| S5.-50.30 | pink T-10, 30 s room tone at -50, pink T-10 | 150 | +10.0 | 2.0 | 0.5 | 29.2 | 38.7 |
| S5.-50.120 | pink T-10, 120 s room tone at -50, pink T-10 | 240 | +10.0 | 2.0 | 1.0 | 40.9 | 42.9 |
| S5.-45.5 | pink T-10, 5 s room tone at -45, pink T-10 | 125 | +10.0 | 0.0 | 0.5 | 0.0 | 5.7 |
| S5.-45.30 | pink T-10, 30 s room tone at -45, pink T-10 | 150 | +10.0 | 2.0 | 0.3 | 25.1 | 31.3 |
| S5.-45.120 | pink T-10, 120 s room tone at -45, pink T-10 | 240 | +10.0 | 2.0 | 0.8 | 35.8 | 42.0 |
| S6.1 | pink T-15, 10 s silence, bursts at T | 130 | ‑0.0 | 3.8 | 1.8 | 17.9 | 28.7 |
| S6.2 | pink T-15, 10 s room tone at -50, bursts at T | 130 | ‑0.0 | 3.8 | 2.8 | 16.9 | 31.5 |
| S8.-60.30 | 30 s room tone at -60, then pink T | 30 | +0.0 | 7.4 | 4.8 | 23.5 | 23.6 |
| S8.-60.120 | 120 s room tone at -60, then pink T | 120 | +0.0 | 7.4 | 6.0 | 38.6 | 141.1 |
| S8.-60.300 | 300 s room tone at -60, then pink T | 300 | +0.0 | 7.4 | 6.0 | 38.6 | 141.1 |
| S8.-50.30 | 30 s room tone at -50, then pink T | 30 | +0.0 | 7.4 | 4.3 | 18.1 | 88.1 |
| S8.-50.120 | 120 s room tone at -50, then pink T | 120 | +0.0 | 7.4 | 5.9 | 23.6 | 24.1 |
| S8.-50.300 | 300 s room tone at -50, then pink T | 300 | +0.0 | 7.4 | 5.9 | 23.6 | 24.1 |
| S8.-45.30 | 30 s room tone at -45, then pink T | 30 | +0.0 | 7.4 | 4.3 | 16.7 | 82.2 |
| S8.-45.120 | 120 s room tone at -45, then pink T | 120 | +0.0 | 7.4 | 5.5 | 19.8 | 90.1 |
| S8.-45.300 | 300 s room tone at -45, then pink T | 300 | +0.0 | 7.4 | 5.5 | 19.8 | 90.0 |
Green: output short-term loudness, with the ±1 LU band. Amber: the gain applied, against the gain a perfect fixed gain would apply (dashed).
S1.+12 A 12 LU step up: too loud for about 40 s, then settled.S1.-12 A 12 LU step down: the lift arrives over about a minute.S2.20 Loud material after a long quiet stretch: the gain arrives on the +12 dB maximum.S3.up The same 12 dB step on programme material.S4.L.30 30 s of digital silence in material that needs a lift.S5.-50.120 Two minutes of room tone at -50 LUFS in a pause.S6.1 Noise bursts after a quiet passage and 10 s of silence.S8.-50.120 The stream starts on two minutes of room tone.Every decision is taken on a 10 ms grid counted from the start of the stream, so the output does not depend on how the host cuts the stream into calls. Checked by a 64-bit hash of every output sample, per 100 ms. The last row is the tuned lab chain the release was ported from.
| Block size | Cases identical | Thread time against 10 ms | Differences |
|---|---|---|---|
| 20 ms | 108 of 108 | 0.99x | every output sample identical |
| 64 frames | 49 of 49 | 1.89x | every output sample identical |
| 256 frames | 108 of 108 | 1.18x | every output sample identical |
| 512 frames | 108 of 108 | 1.09x | every output sample identical |
| 1024 frames | 108 of 108 | 1.04x | every output sample identical |
| 2048 frames | 107 of 108 | 1.01x | S7: from 120.0 s, 85 of 3000 ticks |
| the tuned lab chain | 45 of 45 | 1.00x | every output sample identical |
The cases longer than 16 minutes, with the default (the whole stream) and with a 900 s program length. Neutral on an hour of material; on three hours and more it keeps the long-term level on target.
| Case | Material | Length | Integrated LUFS, default / 900 s | Long-term in band % | W300 in band % |
|---|---|---|---|---|---|
| P2a | Zapping stream, recordings 1-11 | 54 min 58 s | ‑22.80 / ‑22.82 | 99.1 / 99.1 | 98.5 / 99.5 |
| P7 | Poetry, twenty readings in twelve languages | 40 min 23 s | ‑23.05 / ‑23.07 | 99.5 / 99.5 | 100.0 / 100.0 |
| P8.2 | Audiobook, The Cat of Maurice | 39 min 50 s | ‑23.21 / ‑23.17 | 99.4 / 99.4 | 100.0 / 100.0 |
| P10.1 | Music mix 1, classical | 1 h 00 min | ‑22.85 / ‑22.79 | 98.8 / 98.8 | 90.6 / 90.9 |
| P10.2 | Music mix 2, chart pop | 1 h 03 min | ‑22.53 / ‑22.50 | 99.7 / 99.7 | 95.2 / 94.2 |
| P10.3 | Music mix 3, pop | 1 h 10 min | ‑22.36 / ‑22.39 | 100.0 / 100.0 | 89.7 / 94.1 |
| P10.4 | Music mix 4, rock and pop | 31 min 34 s | ‑22.47 / ‑22.48 | 100.0 / 100.0 | 100.0 / 100.0 |
| P10.5 | Music mix 5, workout | 1 h 00 min | ‑22.30 / ‑22.29 | 99.5 / 99.5 | 99.7 / 99.6 |
| P11 | Mixed 1, monologue over an intermittent bed | 59 min 33 s | ‑22.74 / ‑22.75 | 97.5 / 97.5 | 98.8 / 98.7 |
| P12 | Live stream 1, game audio and voice | 30 min 36 s | ‑22.87 / ‑22.89 | 87.0 / 87.0 | 91.0 / 91.0 |
| P13.1 | Conversation 1, European Commission midday briefing (CC BY 4.0) | 49 min 48 s | ‑23.09 / ‑23.10 | 99.9 / 99.9 | 100.0 / 100.0 |
| P13.2 | Conversation 2, FOMC press conference (public domain) | 28 min 42 s | ‑23.14 / ‑23.12 | 99.9 / 99.9 | 100.0 / 100.0 |
| P13.3 | Conversation 3, panel discussion, TIB AV-Portal (CC BY 3.0 DE) | 1 h 24 min | ‑23.41 / ‑23.43 | 96.8 / 96.7 | 84.7 / 85.8 |
| P13.4 | Conversation 4, US Supreme Court oral argument (public domain) | 1 h 49 min | ‑22.70 / ‑22.68 | 99.8 / 99.8 | 100.0 / 100.0 |
| P6 | Long run, four mixes back to back (workout, mixed, pop, rock) | 3 h 42 min | ‑22.45 / ‑22.45 | 92.2 / 99.9 | 85.2 / 94.5 |
P6 Long-term loudness over 3 h 42 min: default and 900 s program length, with the ±1 LU band.The programme cases again at the streaming target. The synthetic cases stay at -23 LUFS, where their levels are defined.
| Pass line | Scope | Result |
|---|---|---|
| G1 | Whole-case integrated output within ±1 LU of ‑16 LUFS | 63 of 67 |
| G2 | Long-term loudness in band, cases longer than 16 min | 11 of 14 |
| G3 | Maximum true peak at or under ‑0.8 dBTP on the EBU / ITU programme files | 4 of 4; highest ‑0.88 dBTP |
| Line | Case | Material | Measured | What it is |
|---|---|---|---|---|
| G1 | P7.02 | Poetry, French 2 | ‑17.65 LUFS | Needs +12.4 dB. Near or beyond the +12 dB the adjuster adds at most, so the target is reached slowly or not at all. |
| G1 | P7.03 | Poetry, German 1 | ‑17.28 LUFS | A 132 s reading that needs a large lift: it ends before the ride has settled. |
| G1 | P9.02 | Big Buck Bunny (Blender, CC-BY) | ‑17.39 LUFS | A quiet film with loud peaks: the true-peak ceiling holds it back, not the boost maximum. |
| G1 | P13.3 | Conversation 3, panel discussion, TIB AV-Portal (CC BY 3.0 DE) | ‑18.79 LUFS | Needs +13.0 dB. Near or beyond the +12 dB the adjuster adds at most, so the target is reached slowly or not at all. |
| G2 | P8.2 | Audiobook, The Cat of Maurice | 89.73 % in band | Needs +11.2 dB. Near or beyond the +12 dB the adjuster adds at most, so the target is reached slowly or not at all. |
| G2 | P12 | Live stream 1, game audio and voice | 50.16 % in band | The sparse live stream again, further under a higher target. |
| G2 | P13.3 | Conversation 3, panel discussion, TIB AV-Portal (CC BY 3.0 DE) | 0.11 % in band | Needs +13.0 dB. Near or beyond the +12 dB the adjuster adds at most, so the target is reached slowly or not at all. |
Everything but the synthetic signals and the EBU / ITU files came through lossy sources, so it is kept out of the true-peak grading (G3): a codec's own inter-sample overshoot cannot be told apart from the device's.