Welcome to the official Sonodit Audio Pipeline documentation. Our platform takes raw voiceover files and transforms them into broadcast-ready, mastered assets through a 100% autonomous process. Below, we detail the chronological journey your audio takes within our intelligent engine:

Step 1: Ingestion and Acoustic Profiling

As soon as the file enters the system, the 'Sonodit Audio Profiler' performs a full diagnostic of the signal. It measures the sample rate, bit depth, and calculates the initial Crest Factor. Simultaneously, a neural network maps the speaker’s phonetic output to determine their exact Words Per Minute (WPM), native language, and regional dialect. This step generates a predictive roadmap that guides the subsequent processing tools.

Step 2: Subsonic Filtering and Noise Subtraction

To ensure maximum clarity, the audio enters the restoration phase. A dynamic High-Pass Filter is set to 50 Hz to eradicate parasitic frequencies muddying the low end. At the same time, the engine captures the 'noise DNA' detected during the speaker's pauses and applies a 50% inverse spectral subtraction algorithm, effectively removing room hiss without altering the natural timbre of the voice.

Step 3: Clip Gain Matching

One of our most sophisticated stages: the audio undergoes dual-stage normalisation. First, the entire signal is standardised to -25.0 dB RMS. Immediately after, the system analyses the audio phrase-by-phrase. Whether the speaker whispered a line or shouted a slogan, the algorithm calculates the exact volume variance and corrects it independently, applying a +/-15.0 dB safety limit while protecting short pauses with an exclusion window. This ensures every word has consistent presence and commercial impact.

Step 4: Dynamic Isolation (Predictive Gating)

The stabilised audio passes through an ultra-precise noise gate configured with a threshold between -50 dBFS and -55 dBFS. The system detects the voice’s attack transients to open the signal flow instantly, applying a controlled fade on word tails to bury any unwanted breath or room ambience into absolute silence.

Step 5: Acoustic Modelling and Commercial Texturisation

The cleaned file is automatically injected into our virtualised DSP engine. Here, the audio is processed through a simulated analogue rack that applies soft-knee compression, strict sibilance control (De-esser), and parallel harmonic excitation. This phase glues the signal together and grants the voice that 'fat and clear' tone characteristic of top-tier radio networks and cinematic productions, with a strict True Peak ceiling of -2.0 dB.

Step 6: Precision Segmentation and PTK Shielding

In the final stage, the temporal orchestrator (Cartographer) takes control. If the system is working autonomously, it locates silent gaps to perform perfect cuts, exporting phrases as individual files. Each cut includes a 10-millisecond preventative attack buffer (PTK Protection), ensuring no leading consonants are clipped. If the user requires a unified master file, the system assembles the phrases by injecting dynamic silences calculated from the speaker's cadence, resulting in a compact, rhythmic piece ready for airplay.