AudioGraph

Latest revision as of 18:04, 16 May 2023

[edit] Description

This filter displays the audio waveform for a video, superimposed on the video. It is mainly intended to help during editing rather than for final output. It can be useful for finding and isolating specific sequences of dialogue or sound, and for checking that overdubbed audio (especially speech) is in sync with video.

The audio is displayed as a green waveform stretching from left to right across the frame. The filter can graph the audio for the currently visible frame only; or it can include the audio for several successive frames on either side of the current frame. Graphing several frames makes it easier to find a sound of interest. It is also really cool to watch the waveform scrolling across the video as the video plays :-).

[edit] Requirements

• [x86]: AviSynth+ or AviSynth 2.6
• [x64]: AviSynth+
• Supported color formats: RGB24, RGB32, YUY2, YV24, 8-bit Planar RGB

[edit] Syntax and Parameters

AudioGraph (clip, int "frames_either_side", int "graph_scale", int "middle_colour", int "side_colour")

clip   =

The source clip. YUY2, YV24, RGB24, RGB32 or planar RGB video, with 8-bit or 16-bit mono or stereo audio. Float audio will be internally converted to 16-bit and back to float.

int  frames_either_side = 25

frames_either_side: The number of frames, either side of the current frame, which should be graphed.

The effect of the frames_either_side parameter is perhaps better explained by this table:

value effect:

• 0 : only audio for the currently visible frame is graphed.
• 1 : audio for the preceding, current, and following frames are graphed.
• 2 : audio for the preceding 2 frames, current frame, and following 2 frames are graphed.
• ...and so on.

The current frame's audio is displayed in the centre of the video frame in bright green, while audio for preceding and following frames is displayed in darker green.

int  graph_scale = 0

graph_scale: The vertical scale factor. Set to 0 to enable auto-scale

int  middle_colour = $00FF00

middle_colour: The graph colour for the current frame. Colors are expressed as hex values: Color_presets

int  side_colour = $7F7F7F

side_colour: The graph colour for the frames on either side of the current. Colors are expressed as hex values: Color_presets

[edit] How this filter works:

An "audioframe" is the audio data corresponding to a video frame, converted into an internal form that can be quickly drawn. A total of (1 + 2 * frames_either_side) audioframes are drawn onto each video frame. Each audioframe is thus (video frame width) / (1 + 2 * frames_either_side) pixels wide. An audioframe simply consists of a Y pixel coordinate for each X pixel coordinate, so drawing an audioframe is very fast.

When frames_either_side is nonzero, the same audioframe will be drawn several times on several successive video frames. So it makes sense to cache audioframes. The filter uses a cache of "audioframe buffers" to store recently used audioframes. Audioframes are generated from raw audio data on demand, and stored in the cache. The caching system is such that a specific audioframe is only ever cached in a specific audioframe buffer, so that cache lookup is very fast. This caching also improves performance when seeking back and forth in a video.

• The included source code is very well documented, check it out for full details!

[edit] Examples

The following script creates a video from just an audio file. The video length is automatically adjusted based on the duration of the audio and the desired frame rate. The output clip will be RGB32, 512x256, 59.94 FPS, and frames_either_side=5, of course this is all adjustable.

WavSource("sample.wav") # for other types of files use FFAudioSource or LWLibavAudioSource or another suitable audio source filter
audio = last

fps = 60000
den = 1001
length = Round((AudioDuration(audio)*fps)/den)
BlankClip(length, width=512, height=256, fps=fps, fps_denominator=den, pixel_type="RGB32", color=$000000)
AudioDub(last, audio)

AudioGraph(frames_either_side=20, graph_scale=0, middle_colour=$00FF00, side_colour=$7F7F7F) # change to desired values

[edit] Changelog

Version       Date            Changes

v0.0.2        2023/05/16      - Update by Asd-g
                              - Fixed undefined behavior - uninitialized optional parameters.
                              - Fixed undefined behavior - uninitialized allocated memory.
                              - Fixed undefined behavior (float input audio) (CPUs with SSE/SSE2) 
                                by reenabling the internal conversion from/to audio float.
                              - Fixed upside down graph. (YUY2)
                              - Added support for YV24 and planar RGB/RGBA (8-bit).
                              - Added support for frame properties passthrough.

v0.0.1        2020/06/21      - Update by Groucho2004
                              - Added version resource
                              - Provided default values for:
                                * frames_either_side (25),
                                * graph_scale (0),
                                * middle_colour ($00FF00),
                                * side_colour ($7F7F7F)
                              - Please note that the parameters now don't have a underscore prefix.

v0.0.0        2020/06/20      - Update by Groucho2004
                              - Update to AviSynth 2.6 api
                              - x86 and x64 binaries

20080928      2008/09/28      - Update by Ihor Bobalo
                              - Includes new parameters "_graph_scale", "_middle_colour", and "_side_colour"	
                              - Only source code provided

20040318      03/18/2004      - AviSynth 2.5 version by Sh0dan
                              - No YV12 support.
                              - Should support multiple channels.
                              - YUY2 mode made a bit more eyepleasing, IMO.
                              - - It makes the graph a bit more blocky.
                              - - YUY2 mode is converted to greyscale.

20030808      08/08/2003      - initial release; AviSynth 2.0 plugin by Richard Ling

[edit] Archived Downloads

[edit] TO DO

The update by Ihor Bobalo includes some of the following features in the to-do list.

• Allow the colour of the graph to be passed as a parameter.
• Allow separate graphing of left or right channels of stereo audio (using different colours).
• Allow a vertical scale factor to be passed as a parameter, so that quiet waveforms can be seen in more detail.
• Fix the "feature" that the current frame's audio is not always centered on the display. It can be offset quite far to the right, depending on the relationship between video frame width and frames_either_side.

[edit] External Links

• avisynth.org.ru - Original documentation, Russian documentation
• SourceForge - AudioGraph update by Ihor Bobalo; only source code provided.

Back to External Filters ←