A transparent work-product generator for speech audio that has to be understood, not beautified: noisy evidence recordings, police/procedural interviews, field footage audio. Following the SWGDE technical notes on FFmpeg for forensic audio examination, the tool never touches your original - it renders an enhanced derivative through a documented, reproducible FFmpeg filter chain that is recorded verbatim in the result metadata, so the processing history of the work product can be reported alongside the exhibit. Four chain presets: EVIDENCE CLARITY (the conservative default) keeps the full speech band and applies FFT denoising (afftdn with a chosen reduction and noise floor), de-essing, gentle 2:1 leveling and speech normalization with a true-peak-safe limiter; INTERVIEW BROADCAST adds a 100 Hz-8 kHz wideband speech window with lighter denoising for publishable interview audio; TELEPHONE BAND band-limits to the classic 300-3400 Hz telephony passband - the range classic intelligibility research standardized - plus a noise gate for pauses; TRANSCRIPTION PREP runs the strongest denoise and gating to feed automatic speech recognition, where consistent level matters more than naturalness. Strength (light / standard / strong) scales the denoise amount and noise-floor setting together, so the chain stays balanced. The report measures the input and output with a BS.1770-4 meter (integrated loudness, true peak) and estimates the noise floor before and after from the quietest 10 percent of 100 ms blocks, so the denoise effect is a number, not a claim. Every enhancement is a processing artifact on a copy: keep the original, document the chain, and disclose the enhancement in your work product - this tool makes the disclosure copy-pasteable.