NAMFORUM Sysop: Created page

2026-09-01T05:52:06Z

Created page

New page

= Creating a Capture =

Creating a reliable capture is a controlled measurement process.

The exact procedure differs between [[Neural Amp Modeler]], [[TONEX]], [[Kemper Profiling]], [[Neural Capture]], [[Line 6 Proxy]], and other systems, but the underlying principles are similar: establish a known signal path, configure the equipment being captured, send the required test or training signal through it, record or measure the response, and evaluate the resulting model.

This article describes the general workflow. Always follow the platform-specific instructions for the capture system being used.

== 1. Decide what you are capturing ==

Before connecting equipment, define exactly what the resulting model is intended to represent.

Possible capture targets include:

* A pedal
* A preamp
* An amplifier without a cabinet
* An amplifier and cabinet
* An amplifier, cabinet, and microphone
* Multiple pedals or processors
* A complete signal chain

This decision determines both the capture signal path and what additional processing may be required during playback.

For example, a direct amplifier capture may require a cabinet or impulse response during playback, while a capture made through a physical cabinet and microphone may already contain the cabinet and microphone characteristics.

See [[Capture Types]] and [[Cabinets and IRs]].

== 2. Configure the device under test ==

Set the equipment to the configuration you want to capture.

Document relevant controls and settings before beginning.

Depending on the device, these may include:

* Channel
* Gain
* Volume
* Tone controls
* Presence or resonance
* Bright, boost, or voicing switches
* Pedal controls
* Cabinet configuration
* Speaker selection
* Microphone and position
* Power or operating mode
* Other devices in the chain

A capture normally represents the behavior of the equipment at a particular configuration.

If several substantially different sounds are required, it may be better to create several captures rather than expect one capture to represent every possible setting.

== 3. Allow equipment to stabilize ==

Equipment should be in its normal operating condition before critical captures are made.

Tube amplifiers in particular should be allowed to warm up and stabilize.

Also verify that:

* Batteries or power supplies are reliable
* Pedals are operating normally
* Amplifier operating modes will not change
* Noise gates or automatic processes are understood
* Nothing is likely to switch state during the capture

A changing device produces changing measurement data.

== 4. Build the capture signal chain ==

Establish the complete path before running the capture.

A common direct chain is:

'''Capture/training signal → audio interface output → reamp or capture device → device under test → audio interface input → capture system'''

A microphone-based amplifier capture may instead use:

'''Capture/training signal → interface → reamp/capture device → amplifier → cabinet → microphone → preamp/interface → capture system'''

A direct amplifier capture may require a suitable load or dedicated capture device between the amplifier's speaker output and the recording interface.

See [[Capture Signal Chain]].

'''Never connect an amplifier speaker output directly to an ordinary audio-interface input unless the equipment is specifically designed to accept that signal.'''

== 5. Verify routing ==

Before running the capture, verify each part of the signal path individually.

Confirm that:

# The intended output carries the test signal.
# The signal reaches the device under test.
# The device responds normally.
# The return reaches the intended interface input.
# The correct input is being recorded.
# Monitoring does not create a feedback loop.
# No unintended processing is active.

Do not rely only on software meters. Where practical, verify that the physical equipment itself is receiving and returning the expected signal.

== 6. Establish gain staging ==

Check the level at every important stage of the chain.

The objective is to avoid unintended clipping while maintaining a useful signal-to-noise ratio.

Potential overload points include:

* Interface output
* Reamp or capture device
* Pedal or processor stages
* Amplifier stages
* Load/direct-output device
* Microphone preamp
* Interface input
* Digital recording path

Remember that a return signal below 0 dBFS does not prove that an earlier analog stage has not clipped.

See [[Gain Staging and Calibration]].

== 7. Calibrate where required ==

If the capture platform provides a calibration procedure, perform it before making the capture.

Calibration may establish the relationship between:

* Digital test-signal level
* Interface output voltage
* Signal presented to the device under test
* Return level
* Model input or output level

Do not substitute an arbitrary level simply because it produces a good-looking waveform.

A capture can sound convincing while still having incorrect level calibration.

Platform-specific calibration requirements should take precedence over generic advice.

== 8. Remove unintended processing ==

Unless specifically required by the capture system, avoid modifying the test or return signal.

Check for:

* EQ
* Compression
* Limiting
* Noise reduction
* Normalization
* Cabinet simulation
* Speaker simulation
* DAW plugins
* Interface DSP
* Automatic gain control
* Sample-rate conversion
* Unintended fades

Any processing included in the measurement path can become part of the behavior the capture system attempts to reproduce.

If processing is intentionally part of the target signal chain, document it.

== 9. Run the capture or record the training response ==

Follow the procedure required by the chosen platform.

Some systems perform the measurement and model creation as part of an integrated capture process.

Other systems, including some [[Neural Amp Modeler|NAM]] workflows, may use a training signal that is played through the device and recorded. The resulting recording is then used in a separate training process.

During the measurement:

* Do not change controls.
* Do not switch channels.
* Avoid touching cables unnecessarily.
* Avoid introducing unrelated sounds into microphone-based captures.
* Watch for clipping or unexpected level changes.
* Allow the process to complete normally.

If the procedure fails, identify the cause before repeatedly running additional captures.

== 10. Preserve the original recording when applicable ==

When a platform uses a recorded training response, retain an unprocessed copy of the original recording.

Do not permanently overwrite the only copy with:

* Normalization
* EQ
* Noise reduction
* Resampling
* Trimming
* Format conversion

If editing is required by the training procedure, work from a copy.

The original recording can be valuable for troubleshooting, retraining, comparison, or future training methods.

== 11. Create or train the model ==

Once the measurement or recording is complete, the capture system creates or trains the resulting model.

Depending on the technology, this may occur:

* Locally on a computer
* Inside dedicated hardware
* Through manufacturer software
* Through an online or cloud service
* Through third-party training software

Training time can vary substantially according to the platform, model architecture, computer hardware, and selected quality settings.

A process completing successfully means that a model was created. It does not necessarily mean that the model is accurate.

== 12. Compare the capture with the original ==

Evaluation is an essential part of the process.

Where practical, compare the model and the original device using the same source signal.

Listen for differences in:

* Overall gain
* Distortion character
* Dynamic response
* Attack
* Compression
* Low-frequency behavior
* High-frequency behavior
* Noise
* Sustain
* Response to different playing levels

Level-match the comparison carefully. A louder signal is easily perceived as more impressive even when it is not more accurate.

For critical evaluation, switching between the original and the model while using the same source material can reveal differences that are difficult to identify from memory.

== 13. Test more than one source ==

A model that sounds accurate with the training or comparison material may behave differently with another source.

Where appropriate, test with:

* Different guitars
* Different pickup types
* Bass
* Different playing dynamics
* Single notes
* Chords
* Clean playing
* Hard transients
* Sustained notes

For captures intended for non-guitar sources, test them with representative material from those sources.

The objective is to determine whether the model reproduces the useful behavior of the original equipment, not merely whether one example sounds similar.

== 14. Name the capture meaningfully ==

A large collection quickly becomes difficult to manage if captures have ambiguous names.

A useful naming convention may identify:

* Manufacturer
* Device
* Channel or mode
* Gain setting
* Cabinet status
* Microphone where applicable
* Significant pedal or switch settings

For example, a collection might distinguish clearly between several captures of the same amplifier at different gain or channel settings.

Avoid names that only make sense to the person who created them.

== 15. Document the capture ==

Good metadata increases the value of a capture.

Useful information can include:

* Equipment manufacturer and model
* Device settings
* Capture date
* Capture platform and software version
* Audio interface
* Reamp or capture device
* Calibration information
* Cabinet and speaker
* Microphone and position
* Microphone preamp
* Sample rate
* Signal-chain details
* Whether a cabinet is included
* Known limitations or unusual behavior

If the capture is shared publicly, this information helps other users understand what they are loading.

== 16. Keep unsuccessful captures ==

Not every unsuccessful experiment should immediately be deleted.

When practical, preserve enough information to understand why a capture failed or produced an unexpected result.

A failed capture can reveal useful information about:

* Calibration
* Signal routing
* Equipment behavior
* Training settings
* Platform limitations
* Noise
* Clipping
* Model architecture

Controlled failures can be useful experiments.

== 17. Share reproducible information ==

When discussing a capture with other users, provide enough information for the problem or result to be understood.

Instead of simply reporting:

''"This capture sounds wrong."''

provide information such as:

* What was captured
* Signal chain
* Interface
* Reamp method
* Calibration
* Device settings
* Capture platform and version
* Playback system
* What differs from the original

This turns an individual experience into information that others can investigate and learn from.

== A repeatable workflow ==

A useful general workflow is:

# Define the capture target.
# Document the equipment settings.
# Build the signal chain.
# Verify routing.
# Establish gain staging.
# Calibrate where required.
# Run the capture.
# Preserve the source measurement where applicable.
# Create or train the model.
# Compare it with the original.
# Test it with varied material.
# Document the result.
# Repeat only the variables you intend to change.

The objective is not merely to produce a model file.

The objective is to produce a model whose origin, behavior, and limitations are understood.

== See also ==

* [[Getting Started with Capture Technology]]
* [[Capture Signal Chain]]
* [[Gain Staging and Calibration]]
* [[Reamping for Capture]]
* [[Capture Types]]
* [[Cabinets and IRs]]
* [[Troubleshooting Captures]]
* [[Glossary]]

'''[https://namforum.com/ Discuss capture techniques and results on NAMFORUM]'''

[[Category:Capture technology]]
[[Category:Capture techniques]]

Creating a Capture - Revision history

NAMFORUM Sysop: Created page