Normalize
From Avisynth wiki
(Difference between revisions)
Raffriff42 (Talk | contribs) m (one more touch-up) |
(add link to avs+ documentation) |
||
| (One intermediate revision by one user not shown) | |||
| Line 1: | Line 1: | ||
<div style="max-width:62em" > | <div style="max-width:62em" > | ||
| + | <div {{BlueBox2|40|0|3px solid purple}} > | ||
| + | {{AvsPlusFullname}}<br> | ||
| + | Up-to-date documentation: [https://avisynthplus.readthedocs.io/en/latest/avisynthdoc/corefilters/normalize.html https://avisynthplus.readthedocs.io] | ||
| + | </div> | ||
| + | |||
| + | |||
Raises (or lowers) the loudest peak of the audio track to a given {{FuncArg|volume}}. This process is called [[Wikipedia:Audio_normalization|''audio normalization'']].<br> | Raises (or lowers) the loudest peak of the audio track to a given {{FuncArg|volume}}. This process is called [[Wikipedia:Audio_normalization|''audio normalization'']].<br> | ||
Note that '''Normalize''' performs [[Wikipedia:Audio_normalization#Peak_normalization|''peak normalization'']] (used to prevent audio [[Wikipedia:Clipping_(audio)#Digital_clipping|clipping]]) and not [[Wikipedia:Audio_normalization#Loudness_normalization|loudness normalization]]. | Note that '''Normalize''' performs [[Wikipedia:Audio_normalization#Peak_normalization|''peak normalization'']] (used to prevent audio [[Wikipedia:Clipping_(audio)#Digital_clipping|clipping]]) and not [[Wikipedia:Audio_normalization#Loudness_normalization|loudness normalization]]. | ||
| Line 10: | Line 16: | ||
:{{Par2|clip |clip|}} | :{{Par2|clip |clip|}} | ||
::Source clip. Supported [[ConvertAudio|audio sample types]]: 16-bit integer and [[Float|32-bit floating-point]]. | ::Source clip. Supported [[ConvertAudio|audio sample types]]: 16-bit integer and [[Float|32-bit floating-point]]. | ||
| − | ::Other sample types (8-, 24- and 32-bit integer) are converted to floating-point | + | ::Other sample types (8-, 24- and 32-bit integer) are automatically converted to floating-point. |
:{{Par2|volume|float|1.0}} | :{{Par2|volume|float|1.0}} | ||
| − | ::Set the amplitude of the loudest audio | + | ::Set the amplitude of the loudest audio. Default = 1.0 for peaking at 0[[Wikipedia:DBFS|dB]]: for [[Float|floating-point]] samples, this corresponds to the range -1.0 to +1.0, and for 16-bit integer samples, this corresponds to the range -32768 to +32767 – the widest range possible without [[Wikipedia:Clipping_(audio)#Digital_clipping|clipping]]. |
::*For a particular peak [[Wikipedia:Decibel|decibel]] level, use the equation {{FuncArg|volume}} = {{Serif|10}}<sup> {{Serif|'''dB''' / 20}}</sup> | ::*For a particular peak [[Wikipedia:Decibel|decibel]] level, use the equation {{FuncArg|volume}} = {{Serif|10}}<sup> {{Serif|'''dB''' / 20}}</sup> | ||
::*For example, set a -3dB peak with {{FuncArg|volume}} = 10<sup>-3/20</sup> or 0.7079. | ::*For example, set a -3dB peak with {{FuncArg|volume}} = 10<sup>-3/20</sup> or 0.7079. | ||
| Line 22: | Line 28: | ||
==== ''Normalization and Floating-point Audio'' ==== | ==== ''Normalization and Floating-point Audio'' ==== | ||
| − | The idea of digital ''clipping'' (when the signal is outside the range that can be stored accurately) really applies only to ''integer'' sample types; floating-point | + | The idea of digital ''clipping'' (when the signal is outside the range that can be stored accurately) really applies only to ''integer'' sample types; floating-point samples will never become clipped in practice, as [[Wikipedia:Single-precision_floating-point_format|the maximum value]] is around 3.4×10<sup>38</sup> – some 29 orders of magnitude (580 dB) larger than 16-bit samples can store. |
| − | '''Normalize''' is therefore not needed for floating-point audio, but using it is recommended before [[ConvertAudio|converting]] to an integer type, especially if any processing has been done | + | '''Normalize''' is therefore not needed for floating-point audio, but using it is recommended before [[ConvertAudio|converting]] to an integer type, especially if any processing has been done – such as [[Amplify|amplification]], [[MixAudio|mixing]] or [[SuperEQ|equalization]] – which may expand the audio peaks beyond the integer clipping range. |
==== Examples ==== | ==== Examples ==== | ||
| Line 57: | Line 63: | ||
</div> | </div> | ||
:[[File:NormalizeEx2_v1,0.png]] | :[[File:NormalizeEx2_v1,0.png]] | ||
| − | :(showing frame 2744 | + | :(showing frame 2744 where the loudest peak was detected, but note that ''Amplify Factor'' is the same for all frames) |
Latest revision as of 05:33, 18 September 2022
AviSynth+
Up-to-date documentation: https://avisynthplus.readthedocs.io
Raises (or lowers) the loudest peak of the audio track to a given volume. This process is called audio normalization.
Note that Normalize performs peak normalization (used to prevent audio clipping) and not loudness normalization.
[edit] Syntax and Parameters
Normalize(clip clip [, float volume , bool show ] )
- clip clip =
- Source clip. Supported audio sample types: 16-bit integer and 32-bit floating-point.
- Other sample types (8-, 24- and 32-bit integer) are automatically converted to floating-point.
- float volume = 1.0
- Set the amplitude of the loudest audio. Default = 1.0 for peaking at 0dB: for floating-point samples, this corresponds to the range -1.0 to +1.0, and for 16-bit integer samples, this corresponds to the range -32768 to +32767 – the widest range possible without clipping.
- For a particular peak decibel level, use the equation volume = 10 dB / 20
- For example, set a -3dB peak with volume = 10-3/20 or 0.7079.
- Where multiple audio channels are present, all channel gains are set in proportion. For example, if the loudest peak on the loudest channel comes to -10dB, by default a gain of +10dB is applied to all channels.
- Set the amplitude of the loudest audio. Default = 1.0 for peaking at 0dB: for floating-point samples, this corresponds to the range -1.0 to +1.0, and for 16-bit integer samples, this corresponds to the range -32768 to +32767 – the widest range possible without clipping.
- bool show = false
- If true, a text overlay (see image below) will show the calculated amplification factor and the frame number of the loudest peak.
[edit] Normalization and Floating-point Audio
The idea of digital clipping (when the signal is outside the range that can be stored accurately) really applies only to integer sample types; floating-point samples will never become clipped in practice, as the maximum value is around 3.4×1038 – some 29 orders of magnitude (580 dB) larger than 16-bit samples can store.
Normalize is therefore not needed for floating-point audio, but using it is recommended before converting to an integer type, especially if any processing has been done – such as amplification, mixing or equalization – which may expand the audio peaks beyond the integer clipping range.
[edit] Examples
- Normalize signal to 98%
video = AviSource("video.avi") audio = WavSource("audio.wav").Normalize(0.98) return AudioDub(video, audio)
- Normalize each channel separately (eg for separate language tracks)
video = AviSource("video.avi") audio = WavSource("audio2ch.wav") left_ch = GetChannel(audio,1).Normalize right_ch = GetChannel(audio,2).Normalize return AudioDub(video, MergeChannels(left_ch, right_ch))
- Effect of show=true with added Histogram, Waveform and current_frame overlays
LoadPlugin(p + "Waveform\waveform.dll") V=BlankClip(pixel_type="YV12", width=480, height=360).Loop A=WavSource("music.wav") AudioDub(V, A).AudioTrim(0.0, A.AudioDuration) ScriptClip(Last, \ """Subtitle(Last, "frame "+String(current_frame), align=5)""") Normalize(volume=1.0, show=true) Histogram(mode="audiolevels") Waveform(window=3) return Last
- (showing frame 2744 where the loudest peak was detected, but note that Amplify Factor is the same for all frames)
