DAW Integration - Plugdata

Overview

plugindata provides deep integration with digital audio workstations (DAWs) through industry-standard plugin formats. This allows Pure Data patches to work seamlessly within modern music production workflows.

Plugin Formats

plugindata is available in multiple plugin formats:

VST3 - Steinberg’s cross-platform format (Windows, macOS, Linux)
AU - Apple Audio Units (macOS only)
LV2 - Open-source format (Linux primarily)
CLAP - Modern open plugin standard (all platforms)

Each format provides the same core functionality with platform-specific optimizations.

Transport Synchronization

Playhead Information

From PluginProcessor.cpp:1055-1128, plugdata receives comprehensive transport data:

void PluginProcessor::sendPlayhead()
{
    AudioPlayHead const* playhead = getPlayHead();
    if (!playhead) return;
    
    auto infos = playhead->getPosition();
    
    lockAudioThread();
    setThis();
    
    if (infos.hasValue()) {
        // Transport state
        atoms_playhead[0] = infos->getIsPlaying();
        sendMessage("__playhead", "playing", atoms_playhead);
        
        atoms_playhead[0] = infos->getIsRecording();
        sendMessage("__playhead", "recording", atoms_playhead);
        
        // Loop information
        atoms_playhead[0] = infos->getIsLooping();
        auto loopPoints = infos->getLoopPoints();
        if (loopPoints.hasValue()) {
            atoms_playhead.emplace_back(
                static_cast<float>(loopPoints->ppqStart));
            atoms_playhead.emplace_back(
                static_cast<float>(loopPoints->ppqEnd));
        }
        sendMessage("__playhead", "looping", atoms_playhead);
        
        // Tempo
        if (infos->getBpm().hasValue()) {
            atoms_playhead.resize(1);
            atoms_playhead[0] = static_cast<float>(*infos->getBpm());
            sendMessage("__playhead", "bpm", atoms_playhead);
        }
        
        // Time signature
        if (infos->getTimeSignature().hasValue()) {
            atoms_playhead.resize(1);
            atoms_playhead[0] = static_cast<float>(
                infos->getTimeSignature()->numerator);
            atoms_playhead.emplace_back(static_cast<float>(
                infos->getTimeSignature()->denominator));
            sendMessage("__playhead", "timesig", atoms_playhead);
        }
        
        // Position (PPQ, samples, seconds)
        auto ppq = infos->getPpqPosition();
        auto samplesTime = infos->getTimeInSamples();
        auto secondsTime = infos->getTimeInSeconds();
        
        if (ppq.hasValue() || samplesTime.hasValue() || 
            secondsTime.hasValue()) {
            atoms_playhead.resize(3);
            atoms_playhead[0] = ppq.hasValue() ? 
                static_cast<float>(*ppq) : 0.0f;
            atoms_playhead[1] = samplesTime.hasValue() ? 
                static_cast<float>(*samplesTime) : 0.0f;
            atoms_playhead[2] = secondsTime.hasValue() ? 
                static_cast<float>(*secondsTime) : 0.0f;
            sendMessage("__playhead", "position", atoms_playhead);
        }
    }
    
    unlockAudioThread();
}

Accessing Transport Data in Pd

Use receive objects to access playhead information:

[r __playhead playing]   ← 0=stopped, 1=playing
|
[sel 1]
|
[metro 500]  ← Only run when playing

Available Receivers:

Receiver	Output	Description
`[r __playhead playing]`	0 or 1	Play/stop state
`[r __playhead recording]`	0 or 1	Recording state
`[r __playhead looping]`	0/1 start end	Loop state and points (PPQ)
`[r __playhead bpm]`	float	Current tempo
`[r __playhead timesig]`	num denom	Time signature
`[r __playhead position]`	ppq samples seconds	Transport position (list)
`[r __playhead lastbar]`	float	PPQ of last bar start
`[r __playhead framerate]`	float	Frame rate (for video sync)
`[r __playhead edittime]`	float	Edit origin time

Tempo Sync Example

[r __playhead bpm]
|
[/ 60]           ← Convert BPM to beats per second
|
[/ 4]            ← Quarter notes to 16th notes
|
[1000 /]         ← Convert to milliseconds
|
[metro]          ← Tempo-synced 16th note clock

Position-Based Triggering

[r __playhead position]
|
[unpack f f f]
|       |    |
PPQ  Samples Seconds
|
[int]                  ← Get integer PPQ (beats)
|
[change]               ← Trigger on beat change
|
[bang]                 ← Beat trigger

Automation

Parameter System

From PluginProcessor.cpp:120-131, plugdata exposes automation parameters:

// Built-in volume parameter
auto* volumeParameter = new PlugDataParameter(this, "volume", 
    0.8f,   // default value
    true,   // is volume param
    0,      // parameter index
    0.0f,   // minimum
    1.0f    // maximum
);
addParameter(volumeParameter);
volume = volumeParameter->getValuePointer();

// General purpose automation parameters
for (int n = 0; n < numParameters; n++) {
    auto* parameter = new PlugDataParameter(this, 
        "param" + String(n + 1),
        0.0f,    // default value
        false,   // not volume param
        n + 1,   // parameter index
        0.0f,    // minimum
        1.0f     // maximum
    );
    addParameter(parameter);
}

Default Configuration:

1 volume parameter (param0/volume)
64 general automation parameters (param1-param64)
Range: 0.0 to 1.0 (normalized)
Saved with DAW project

Parameter Sending

From PluginProcessor.cpp:1149-1159, parameters are efficiently updated:

void PluginProcessor::sendParameters()
{
    ScopedLock lock(audioLock);
    for (auto* param : enabledParameters) {
        if (EXPECT_UNLIKELY(param->wasChanged())) {
            auto title = param->getTitle();
            sendFloat(title.data(), param->getUnscaledValue());
            param->setUnchanged();
        }
    }
}

Parameters are only sent when they change, minimizing overhead.

Using Automation in Pd

Receive Parameter Values:

[r param1]
|
[* 440]     ← Scale 0-1 to 0-440 Hz
|
[osc~]

Multiple Parameters:

[r param1]  [r param2]  [r param3]
|           |           |
[* 1000]    [* 0.5]     [* 127]
|           |           |
Freq        Resonance   Velocity

Bi-directional Control:

[r param1]     [loadbang]
|              |
[hsl 128 15]   [0.5(
|              |
[s param1]     ← Send back to parameter

Sending values back to parameters from Pd can conflict with DAW automation. Use carefully and consider disabling parameter sending when DAW automation is active.

Parameter Mapping

Linear Scaling:

[r param1]
|
[expr $f1 * (max - min) + min]  ← Scale to range

Exponential Scaling (for frequency):

[r param1]
|
[expr pow(2, $f1 * 10)]  ← Exponential curve
|
[* 20]                   ← Scale to Hz

Stepped Values:

[r param1]
|
[* 12]       ← 0-12 range
|
[int]        ← Round to integer
|
[sel 0 1 2 3 4 5 6 7 8 9 10 11 12]

MIDI Routing

MIDI Input Processing

From PluginProcessor.cpp:1166-1201, MIDI is processed and sent to Pd:

void PluginProcessor::sendMidiBuffer(int const device, 
                                    MidiBuffer const& buffer)
{
    if (acceptsMidi()) {
        for (auto event : buffer) {
            auto message = event.getMessage();
            auto const channel = message.getChannel() + (device << 4);
            
            if (message.isNoteOn()) {
                sendNoteOn(channel, message.getNoteNumber(), 
                          message.getVelocity());
            } else if (message.isNoteOff()) {
                sendNoteOn(channel, message.getNoteNumber(), 0);
            } else if (message.isController()) {
                sendControlChange(channel, 
                    message.getControllerNumber(), 
                    message.getControllerValue());
            } else if (message.isPitchWheel()) {
                sendPitchBend(channel, 
                    message.getPitchWheelValue() - 8192);
            } else if (message.isChannelPressure()) {
                sendAfterTouch(channel, 
                    message.getChannelPressureValue());
            } else if (message.isAftertouch()) {
                sendPolyAfterTouch(channel, 
                    message.getNoteNumber(), 
                    message.getAfterTouchValue());
            } else if (message.isProgramChange()) {
                sendProgramChange(channel, 
                    message.getProgramChangeNumber());
            } else if (message.isSysEx()) {
                for (int i = 0; i < message.getSysExDataSize(); ++i) {
                    sendSysEx(device, message.getSysExData()[i]);
                }
            } else if (message.isMidiClock() || 
                      message.isMidiStart() || 
                      message.isMidiStop() || 
                      message.isMidiContinue()) {
                for (int i = 0; i < message.getRawDataSize(); ++i) {
                    sendSysRealTime(device, message.getRawData()[i]);
                }
            }
            
            // Send raw MIDI bytes
            for (int i = 0; i < message.getRawDataSize(); i++) {
                sendMidiByte(device, message.getRawData()[i]);
            }
        }
    }
}

MIDI in Pure Data

Note Messages:

[notein]
|  |   |
note vel channel

Control Change:

[ctlin]
|  |   |
val cc# channel

Pitch Bend:

[bendin]
|       |
value   channel  ← Range: -8192 to 8191

Complete MIDI Input:

[notein]  [ctlin]  [bendin]  [pgmin]  [touchin]  [polytouchin]

MIDI Output

Send MIDI from Pd back to the DAW:

[60 100(
|  |
[noteout 1]  ← Send to DAW channel 1

Complete MIDI Output:

[noteout]  [ctlout]  [bendout]  [pgmout]  [touchout]  [polytouchout]

Multi-Timbral Setup

DAW MIDI can address 16 channels per device:

[notein]
|  |  |
|  |  [sel 1]  ← Filter for channel 1
|  |  |
[pack f f]
|
[makenote 127 500]  ← Synth voice for channel 1

Sample Rate Handling

Sample Rate Adaptation

From PluginProcessor.cpp:699-759, plugdata adapts to DAW sample rate:

void PluginProcessor::prepareToPlay(double const sampleRate, 
                                   int const samplesPerBlock)
{
    if (approximatelyEqual(sampleRate, 0.0))
        return;
    
    float const oversampleFactor = 1 << oversampling;
    auto maxChannels = std::max(getTotalNumInputChannels(), 
                                getTotalNumOutputChannels());
    
    // Initialize Pd at (possibly oversampled) sample rate
    prepareDSP(getTotalNumInputChannels(), 
              getTotalNumOutputChannels(), 
              sampleRate * oversampleFactor);
    
    // Setup oversampler
    oversampler = std::make_unique<dsp::Oversampling<float>>(
        std::max(1, maxChannels), 
        oversampling, 
        dsp::Oversampling<float>::filterHalfBandPolyphaseIIR, 
        false
    );
    
    oversampler->initProcessing(samplesPerBlock);
    
    statusbarSource->setSampleRate(sampleRate);
    statusbarSource->setBufferSize(samplesPerBlock);
}

Supported Sample Rates:

44.1 kHz (CD quality)
48 kHz (standard digital)
88.2 kHz (2x oversample of 44.1)
96 kHz (2x oversample of 48)
176.4 kHz (4x oversample of 44.1)
192 kHz (4x oversample of 48)
Any DAW-specified rate

Sample Rate in Pd

Access current sample rate:

[samplerate~]
|
[print samplerate]  ← Outputs: 48000 (example)

Use for sample-rate-independent delays:

[samplerate~]
|
[/ 1000]        ← Samples per millisecond
|
[* 10]          ← 10ms delay
|
[vd~ delayline] ← Variable delay

Buffer Size Handling

Block Size Adaptation

Pure Data processes in fixed 64-sample blocks, but DAWs use variable buffer sizes. From PluginProcessor.cpp:724-758:

auto const pdBlockSize = static_cast<size_t>(Instance::getBlockSize());
audioBufferIn.setSize(maxChannels, pdBlockSize);
audioBufferOut.setSize(maxChannels, pdBlockSize);

// If the block size is a multiple of 64 and we are not a plugin, 
// we can optimise the process loop
variableBlockSize = !ProjectInfo::isStandalone || 
                   samplesPerBlock < pdBlockSize || 
                   samplesPerBlock % pdBlockSize != 0;

if (variableBlockSize) {
    // Use FIFO buffers for arbitrary block sizes
    inputFifo = std::make_unique<AudioMidiFifo>(
        maxChannels, 
        std::max<int>(pdBlockSize, samplesPerBlock * oversampleFactor) * 3
    );
    outputFifo = std::make_unique<AudioMidiFifo>(
        maxChannels, 
        std::max<int>(pdBlockSize, samplesPerBlock * oversampleFactor) * 3
    );
    outputFifo->writeSilence(Instance::getBlockSize());
}

Processing Modes

Constant Block Size (optimized):

DAW buffer is multiple of 64 samples
Direct processing without buffering
Lower latency
Used in standalone mode when possible

Variable Block Size (adaptive):

DAW buffer is arbitrary size
FIFO buffering required
Slightly higher latency
Always used in plugin mode

From PluginProcessor.cpp:941-989:

void PluginProcessor::processConstant(dsp::AudioBlock<float> buffer)
{
    int const pdBlockSize = Instance::getBlockSize();
    int const numBlocks = buffer.getNumSamples() / pdBlockSize;
    
    for (int block = 0; block < numBlocks; block++) {
        // Process MIDI
        midiDeviceManager.dequeueMidiInput(pdBlockSize, 
            [this](int const port, MidiBuffer const& buffer) {
                sendMidiBuffer(port, buffer);
            });
        
        // Copy input
        for (int ch = 0; ch < buffer.getNumChannels(); ch++) {
            FloatVectorOperations::copy(
                audioVectorIn.data() + ch * pdBlockSize,
                buffer.getChannelPointer(ch) + audioAdvancement,
                pdBlockSize);
        }
        
        setThis();
        sendParameters();
        sendMessagesFromQueue();
        
        // Process audio (64 samples)
        performDSP(audioVectorIn.data(), audioVectorOut.data());
        
        // Copy output
        for (int ch = 0; ch < buffer.getNumChannels(); ch++) {
            FloatVectorOperations::copy(
                buffer.getChannelPointer(ch) + audioAdvancement,
                audioVectorOut.data() + ch * pdBlockSize,
                pdBlockSize);
        }
        
        audioAdvancement += pdBlockSize;
    }
    
    audioAdvancement = 0;
}

Latency Compensation

From PluginProcessor.cpp:192:

setLatencySamples(pd::Instance::getBlockSize());

plugindata reports 64 samples of algorithmic latency to the DAW, which automatically compensates by:

Delaying other tracks
Adjusting playback timing
Maintaining phase alignment

Additional latency from oversampling is also reported:

setLatencySamples(pd::Instance::getBlockSize() + oversamplerLatency);

Session Management

State Saving

From PluginProcessor.cpp:1224-1308, complete state is saved:

void PluginProcessor::getStateInformation(MemoryBlock& destData)
{
    setThis();
    MemoryOutputStream ostream(destData, false);
    
    ostream.writeInt(patches.size());
    
    lockAudioThread();
    for (auto const& patch : patches) {
        auto content = patch->getCanvasContent();  // Full patch
        auto patchFile = patch->getCurrentFile().getFullPathName();
        
        // Replace absolute paths with variables
        if (patchFile.startsWith(patchesDir.getFullPathName())) {
            patchFile = patchFile.replace(
                patchesDir.getFullPathName(), 
                "${PATCHES_DIR}"
            );
        }
        
        // Write legacy format
        ostream.writeString(content);
        ostream.writeString(patchFile);
    }
    unlockAudioThread();
    
    // Save settings
    ostream.writeInt(getLatencySamples() - Instance::getBlockSize());
    ostream.writeInt(oversampling);
    ostream.writeFloat(getValue<float>(tailLength));
    
    // XML format for extensibility
    auto xml = XmlElement("plugdata_save");
    xml.setAttribute("Version", PLUGDATA_VERSION);
    xml.setAttribute("Oversampling", oversampling);
    xml.setAttribute("Latency", getLatencySamples());
    xml.setAttribute("TailLength", getValue<float>(tailLength));
    xml.setAttribute("Width", editor->getWidth());
    xml.setAttribute("Height", editor->getHeight());
    
    // Save parameter values
    PlugDataParameter::saveStateInformation(xml, getParameters());
    
    // Write XML to stream
    MemoryBlock xmlBlock;
    copyXmlToBinary(xml, xmlBlock);
    ostream.writeInt(static_cast<int>(xmlBlock.getSize()));
    ostream.write(xmlBlock.getData(), xmlBlock.getSize());
}

Saved Data:

Complete patch content (embedded)
Parameter values
Oversampling settings
Editor size and position
Connection paths (segmented)
Custom state data
Plugin version

State Loading

State is restored when DAW project opens:

void PluginProcessor::setStateInformation(void const* data, 
                                         int const sizeInBytes)
{
    MemoryInputStream istream(data, sizeInBytes, false);
    
    int const numPatches = istream.readInt();
    
    SmallArray<std::pair<String, File>> newPatches;
    for (int i = 0; i < numPatches; i++) {
        auto state = istream.readString();  // Patch content
        auto path = istream.readString();   // File path
        
        // Expand path variables
        path = path.replace("${PATCHES_DIR}", 
                          patchesDir.getFullPathName());
        
        newPatches.emplace_back(state, File(path));
    }
    
    // Recreate patches
    for (auto& [content, file] : newPatches) {
        auto* newPatch = loadPatchFromString(content, file);
    }
}

Performance Optimization

CPU Load Management

From PluginProcessor.cpp:744:

cpuLoadMeasurer.reset(sampleRate, samplesPerBlock);

CPU usage is measured and displayed in the status bar.

Audio Thread Safety

From PluginProcessor.cpp:825-938, processing is thread-safe:

void PluginProcessor::processBlock(AudioBuffer<float>& buffer, 
                                  MidiBuffer& midiBuffer)
{
    ScopedNoDenormals noDenormals;  // Prevent denormal slowdown
    AudioProcessLoadMeasurer::ScopedTimer cpuTimer(
        cpuLoadMeasurer, buffer.getNumSamples());
    
    setThis();  // Set Pd instance context
    
    sendPlayhead();  // Update transport info
    
    // Process with oversampling if enabled
    auto targetBlock = dsp::AudioBlock<float>(buffer);
    auto const blockOut = oversampling > 0 ? 
        oversampler->processSamplesUp(targetBlock) : targetBlock;
    
    if (variableBlockSize) {
        processVariable(blockOut, midiBuffer);
    } else {
        midiBuffer.clear();
        processConstant(blockOut);
    }
    
    if (oversampling > 0) {
        oversampler->processSamplesDown(targetBlock);
    }
    
    // Apply volume with smoothing
    smoothedGain.setTargetValue(mappedTargetGain);
    smoothedGain.applyGain(buffer, buffer.getNumSamples());
    
    // Send MIDI output
    midiDeviceManager.sendAndCollectMidiOutput(midiBuffer);
    
    // Optional limiter for safety
    if (enableLimiter && buffer.getNumChannels() > 0) {
        auto block = dsp::AudioBlock<float>(buffer);
        limiter.process(block);
    }
}

Best Practices

Optimizing for DAW Use

Minimize CPU Usage:

Use efficient algorithms
Disable unused signal paths
Consider oversampling cost
Monitor CPU meter

Handle Automation Smoothly:

[r param1]
|
[line~]  ← Smooth parameter changes

Tempo Sync Correctly:

[r __playhead bpm]
|
[change]  ← Only update on tempo change

MIDI Latency:

MIDI is processed per-block (64 samples)
For tighter timing, reduce DAW buffer size
Consider MIDI->audio jitter in timing-critical applications

Testing in DAW

Test automation recording and playback
Verify tempo sync at different BPMs
Check MIDI routing and channel handling
Test state recall (save/load project)
Monitor CPU usage under load
Test multiple instances for efficiency
Verify latency compensation with other tracks

Troubleshooting

Common Issues

No Sound

Check DAW routing to plugin
Verify [dac~] connections in patch
Check output channel configuration
Ensure DSP is enabled (should be automatic)

Automation Not Working

Verify parameter receivers: [r param1]
Check DAW automation is enabled
Test with simple patch first
Look for parameter conflicts

Transport Not Syncing

Check receivers: [r __playhead ...]
Test with simple tempo display
Verify DAW is sending transport
Check if transport is playing

High CPU Usage

Disable oversampling if not needed
Optimize Pd patch efficiency
Check for unnecessary computations
Use [block~] for lower rates where possible

MIDI Not Received

Verify MIDI track routes to plugin
Check MIDI channel filtering
Test with [notein] + [print]
Ensure DAW MIDI output is enabled

Next Steps

Plugin vs Standalone

Understand the differences between operating modes

Patching Basics

Master Pure Data patching fundamentals

​Overview

​Plugin Formats

​Transport Synchronization

​Playhead Information

​Accessing Transport Data in Pd

​Tempo Sync Example

​Position-Based Triggering

​Automation

​Parameter System

​Parameter Sending

​Using Automation in Pd

​Parameter Mapping

​MIDI Routing

​MIDI Input Processing

​MIDI in Pure Data

​MIDI Output

​Multi-Timbral Setup

​Sample Rate Handling

​Sample Rate Adaptation

​Sample Rate in Pd

​Buffer Size Handling

​Block Size Adaptation

​Processing Modes

​Latency Compensation

​Session Management

​State Saving

​State Loading

​Performance Optimization

​CPU Load Management

​Audio Thread Safety

​Best Practices

​Optimizing for DAW Use

​Testing in DAW

​Troubleshooting

​Common Issues

​Next Steps

Plugin vs Standalone

Patching Basics

Overview

Plugin Formats

Transport Synchronization

Playhead Information

Accessing Transport Data in Pd

Tempo Sync Example

Position-Based Triggering

Automation

Parameter System

Parameter Sending

Using Automation in Pd

Parameter Mapping

MIDI Routing

MIDI Input Processing

MIDI in Pure Data

MIDI Output

Multi-Timbral Setup

Sample Rate Handling

Sample Rate Adaptation

Sample Rate in Pd

Buffer Size Handling

Block Size Adaptation

Processing Modes

Latency Compensation

Session Management

State Saving

State Loading

Performance Optimization

CPU Load Management

Audio Thread Safety

Best Practices

Optimizing for DAW Use

Testing in DAW

Troubleshooting

Common Issues

Next Steps