Welcome to the official Sonodit Audio Pipeline documentation. Our platform transforms raw voiceover takes into broadcast-ready, mastered assets—completely autonomously. Here is the chronological journey your audio takes within our intelligent engine:

Step 1: Ingestion and Acoustic Profiling

As soon as your file hits our system, the Sonodit Audio Profiler performs a deep-scan of the signal. It maps the sample rate, bit depth, and calculates the initial Crest Factor. Simultaneously, a neural network analyses the speaker's phonetic map to determine their exact Words Per Minute (WPM), native language, and regional dialect. This creates a predictive roadmap that guides the rest of the processing chain.

Step 2: Subsonic Filtering and Noise Subtraction

To ensure total sonic purity, the audio enters the restoration phase. A dynamic 50Hz High-Pass Filter is applied to scrub away unwanted low-end rumble. At the same time, the engine captures the "noise DNA" detected during natural pauses and applies a 50% inverse spectral subtraction algorithm, wiping out room hiss without compromising the natural timbre of the voice.

Step 3: Block Stabilisation (Clip Gain Matching)

This is one of our most sophisticated phases: dual-stage normalisation. First, the entire signal is levelled to -25.0 dB RMS. Immediately after, the system analyses the audio phrase-by-phrase. Whether the speaker whispered a line or shouted a hook, the algorithm calculates the exact gain discrepancy and corrects it independently. We apply a +/-15.0 dB safety limit, using an exclusion window to protect short pauses, ensuring every word lands with maximum commercial impact.

Step 4: Dynamic Isolation (Predictive Gating)

The stabilised audio then passes through an ultra-precise noise gate, set between -50 dBFS and -55 dBFS. The system detects the voice’s attack transients to open the signal path instantly, applying a controlled fade on word tails to bury any unwanted breaths or air noise in absolute silence.

Step 5: Acoustic Modelling and Commercial Texturising

The cleaned audio is injected into our virtualised DSP Engine. Here, the signal runs through a simulated analogue rack, adding soft-knee compression, surgical de-essing, and parallel harmonic excitation. This phase glues the signal together, giving your voice that classic "fat and clear" radio-ready sheen, with the peak strictly capped at -2.0 dB True Peak.

Step 6: Precision Segmentation and PTK Shielding

In the final stage, our temporal orchestrator (The Cartographer) takes over. For automated workflows, it identifies silence valleys to slice your audio into individual files. Every clip includes a 10ms predictive attack margin (PTK Protection) to ensure no initial consonants are clipped. If you’ve requested a unified master, the system assembles the phrases by injecting dynamic silences calculated from the speaker's rhythm, delivering a tight, punchy, and sales-ready final asset.