The forward method runs a neural model as an audio effect: audio in, neural-transformed audio out

The most direct way to generate sound with a RAVE model through nn~ is the forward method. The object syntax is nn~ MODEL_NAME [METHOD_NAME] [BUFFER_SIZE], where METHOD_NAME defaults to forward and BUFFER_SIZE is the inner buffer nn~ uses to transform sound (smallest by default). Plug an audio signal into the inlet and take the transformed signal from the outlet — the model reconstructs the incoming sound in the timbral character of its training dataset, acting as a neural audio effect. Larger buffer sizes trade latency for stability; forward is the encode-then-decode chain run automatically.

Examples

nn~ vintage.ts forward 2048 — loads the vintage RAVE model, forward method, 2048-sample buffer. Feed a synth voice in; the outlet emits it re-synthesized through the model.

Assessment

Write the nn~ object arguments to run model vintage.ts as an audio effect with a 2048-sample buffer, and describe what appears at the outlet when you feed it live audio.