Blender Audio Documentation
(includes Speaker Object documentation)
There are two properties you can check when loading a sound:
- Cache: This means the whole sound will be decoded on load and the raw audio data will be buffered in memory, which results in faster playback, but uses quite a lot of memory. So this should be used for short sound effects that get played more often, but not for longer audio files such as music.
- Mono: For any 3D Audio or Panning effects the sound source has to be single channel, otherwise it's assumed that the 3D audio and panning information is already present in the multichannel file. So if you want to use those effects for a file with multiple channels, check this!
Sequencer Audio Strips
There are several settings in the properties panel of the Video Sequence Editor:
- Caching: See Cache option on file loading.
- Draw Waveform: This draws the outline of the waveform in the audio strip which is a particularly useful feature for lip synching.
- Volume: Animatable! Controls the volume of the strip. Typically values should be between 0 and 1! If you use higher values it's possible that clipping (samples with an resulting amplitude out of this range get clipped to it) happens, which drastically influences sound quality.
- Pitch: Animatable! This basically changes the playback speed of the sound which also results in a pitch change. Unfortunately this leads to possible seeking errors and the length of the strip isn't updated as well.
- Pan: Animatable! Only works for Mono sources. Values can be between -2 and 2, where 0 means front/center, -1 means to the left and 1 to the right. In case of multichannel audio/rear speakers you can pan to those with the higher values: -2, 2 is back. So this value basically represents the angle at which it's played, to get it in degrees: angle = pan * 90°.
Speaker objects also have animatable volume and pitch settings. Moreover they have 3D audio relevant settings that are not animatable.
Distance attenuation relevant settings
- Minimum Volume: No matter how far the object is away, the distance based volume won't be lower than this value.
- Maximum Volume: No matter how close the object is, the distance based volume won't be higher than this value.
- Attenuation/Rolloff: This factor changes the strength of the distance based volume change, depending on the chosen distance model (See scene settings).
- Maximum Distance: If the object is farther away than this distance, this distance is used to calculate the distance based volume. Influence of this value also depends on the distance model.
- Reference Distance: The distance at which the volume is 100 % (1.0). Set this value to the distance used for recording the sound. Usually sound effects recordings should be made exactly 1 m away from sound to get an accurate volume.
Cone/directionality relevant settings
Imagine a cone with the top at the original of the speaker object and the main axis of it facing in the same direction as the speaker. There are two cones an inner and an outer cone. The angles represent their opening angles, so 360 ° mean the cone is fully open and there's no directionality anymore. Inside the inner cone the volume is 100 % (1.0), outside the outer cone the volume is whatever one sets for the outer cone volume and the volume between those two cones is linearly interpolated between this two volumes.
There's no setting to choose the start time when the speaker should start playing, because you might want a single speaker to play multiple times. Therefore you have to open the NLA editor where you can add sound strips that tell when the sound should start (nothing else, so any other properties of the strips, like length don't matter). When you add a speaker object there's already such a strip added for you at the current frame. Shortcut to add a strip in the NLA Editor: Shift + K.
Those are in the Properties editor -> Scene -> Audio.
Volume Property: Animatable. Overall scene volume that applies for sequencer strips and speaker objects. See Sequencer Audio Strips for details.
Update animation cache: The audio system has an internal cache for the animated values that has to be kept up to date with blender. Normally this is done during every frame change, so if you fully play back the scene once, the cache should be up to date. Sometimes you want this to happen quicker, so just press this button to get an up to date cache.
These settings are relevant for the 3D Audio calculation of speaker objects as well as in the Game Engine.
- Distance Model: Changes how the distance based attenuation is calculated, physically correct would be the Inverse model, but it's also possible to choose a linear and an exponential falloff. The clamped modes mean that the volume can't be louder than 100 % (1.0), that means if the distance is smaller than the reference distance, the volume is always 100 % (1.0).
- Speed of Sound: important for the doppler effect calculation. The typical value is 343.3 m/s in air, in water for example this value is around 1560 m/s. Physically the value depends on the medium, temperature, density and other values.
- Doppler effect: Controls how strong the doppler effect is. This is no physically based value, it's only there so you can exaggerate (> 1.0) or attenuate (< 1.0) the effect, physically correct is a factor of 1.0.
For mixdown you can first set channel count and sample rate for the scene:
- Channel Count: Basically states how many speakers are there for playback later. Typical value is stereo. For home theatres there are often 5.1 or 7.1 sound systems.
- Sample Rate: Tells how many audio samples per second there are. Typical values are 8000 and 11025 Hz and multiples of them. Values lower than 40000 Hz cannot contain all hearable frequencies. So common used values are: 44100 Hz (Music files, Audio CDs), 48000 Hz (recommended for Video files, DAT, DVD, BluRay), 96000 or 192000 (considered very high fidelity, DVD and BluRay).
Then when clicking mixdown you have some more options to set:
- Accuracy: Size of the mixdown buffer that also decides how often the animation relevant data is recalculated. A size of 1 gives 100 % accuracy, but will take a very long time to calculate. 1024 is a good basic value, that you might decrease by a factor of two only if you notice problems in the output.
- Container: File format for the mixdown. Recommended: matroska/mkv: Can save any Codec, or use the fitting Container for the Codec you want to use (eg Ogg for Vorbis), see next point.
- Codec: If the container supports multiple codecs (like matroska/mkv or ogg), you can choose the audio codec. Examples: PCM (Uncompressed RAW data, no quality loss, but huge), FLAC (Free Lossless Audio Codec, Compressed and lossless, recomended for high fidelity), Vorbis (similar to mp3, but free!), AC3 (Attrac-3, Dolby Digital, used on DVD and BluRay).
- Format: Depending on the Codec, Tells how many bits and which format are used for a single sample. Common used standard in audio is S16 (Signed integer, 16 bit), meanings of the letters: S = Signed, U = Unsigned, F = Floating Point. Internally blender uses F32, so this should be your choice for high fidelity, otherwise S16 is fine (and used for all Codecs where you don't have a choice).
In the timeline window you can set playback settings.
- Synchronisation mode: No Sync means that audio plays back realtime and video displays every frame, no matter how fast, resulting in a non synchronised playback, if the video is too slow. Frame Dropping: Again sound plays back realtime, but video plays clocked by an external timer and drops frames if necessary, this should be more precise, but if the audio playback and the external timer are out of sync, playback will also be. AV-sync: Audio plays back realtime and tells video it's playback time resulting in synchronised playback (video drops frames if necessary). If Jack is used as output device, this enables Jack Transport support.
- Scrubbing: When there's no animation played back the sound of the current frame is played whenever the frame changes, providing a simple review (rehear? :D).
Game Engine - Sound Actuator
The Sound Actuator basically has the same properties as the speaker objects have, so check above. Moreover it has a playback mode. End or Stop in the Mode name tell, that for End the sound play until the end of the sample and for Stop it stops immediately when the positive signal is gone. Apart from that addition the modes are: Play (aka play once), Loop (aka loop endlessly) and Loop Bidirectional (aka Play the sample forward, then backward and loop this process endlessly, in case of End this stops as soon as the backward playing is over).
In the User Preferences window under System you find the sound settings:
- Output Device: None (obvious), SDL (Simple direct media layer, should work everywhere, but bad quality), OpenAL (Standard, supports Hardware mixing with correct drivers), Jack (Audio Server for professional audio)
- Channels: Output channels, ie how many speakers do you have? (Only really works for Jack, SDL is buggy and OpenAL autodetects and uses this or some external user configuration)
- Mixing Buffer: Size of the mixing buffer, always has an influence on the latency, so for this the lower the better, but if you have clicks or other problems, try rising the size.
- Sample rate: See Mixdown.
- Sample format: also see Mixdown.
Bake Sound to F-Curves
This operator is available in the Graph/F-Curve Editor: Key -> Bake Sound to F-Curves. It bakes the audio wave hull curve onto an F-Curve. Settings:
- Lowest frequency: Cutoff frequency of a highpass that is applied to the audio data.
- Highest frequency: Cutoff frequency of a lowpass that is applied to the audio data.
- Attack time: Value for the hull curve calculation that tells how fast the hull curve can rise (the lower the value the steeper it can rise).
- Release time: Value for the hull curve calculation that tells how fast the hull curve can fall (the lower the value the steeper it can fall).
- Threshold: Threshold for the hull curve calculation, that thresholds the minimum amplitude value needed to influence the hull curve.
- Accumulate: In case this is on, only the positive differences of the hull curve amplitudes are summarized to produce the output.
- Additive: The amplitudes of the hull curve are summarized or when Accumulate is checked the differences (also negative) are additionally added.
- Square: The output is squared which means the signum function is applied to the hull curve, resulting in a graph with only -1, 0 or 1 as value.
- Square Threshold: Threshold for squaring, all values with an absolute amplitude lower as this value result in 0 as value.