One toolkit bundle for recording, cleaning, inspecting, and preparing speech datasets.

Recursive WAV trimmer for mono files. It reads thresholds and fade settings from a local .env, trims in place, and writes a trim_report.txt sorted by total trimmed milliseconds.

• Put the exe and its .env next to the target folder or pass a folder path.
• Set pre/post buffer and fade values in .env.
• Run it on the folder and read trim_report.txt.

Scans the current folder recursively for WAV files, writes per-file duration lines, and appends the total dataset duration to durations_precise.txt.

• Run it inside the folder you want to measure.
• It scans all WAV files below that folder.
• Open durations_precise.txt for the totals.

Measures the first and last ten-millisecond windows that cross a threshold, then reports the lowest pre, lowest post, and lowest combined values in pre_post_report.txt.

• Set THRESHOLD in .env if needed.
• Run it in the target folder.
• Read pre_post_report.txt for flagged edge cases.

Dataset-structure report tool that groups results by dataset, gender, speaker, language, and emotion, then writes average RMS, min RMS, max RMS, and max peak values to rms_report_by_emotion.txt.

• Use --root to point to your dataset root.
• Optionally set --out and --no-fx.
• Open the generated report grouped by emotion.

Finds the lowest-RMS 32-bit float WAV files in the current tree and writes the sorted list to lowest_overall_rms.txt.

• Run it inside the target folder.
• It scans only valid 32-bit float WAV files.
• Read lowest_overall_rms.txt for the lowest files.

Finds the quietest onset windows using RMS, peak, or both modes and writes report files with onset time and level.

• Set THRESHOLD and MODE in .env.
• Run it in the target folder.
• Check the saved reports for the quietest onsets.

Scans for clipping-like runs of consecutive samples above a threshold and writes the flagged relative paths to a report file.

• Set the input folder, output file, threshold, and consecutive-hit values in .env.
• Run the tool on the target folder.
• Read the saved report for flagged files.

Flags zero-length files, very short files, unsupported formats, very low-energy files, and corrupted WAVs, then writes the results to ras_report.txt.

• Run it inside the folder you want to check.
• It walks subfolders recursively.
• Open ras_report.txt for all flagged files.

Scans 32-bit float WAV files, ranks them by maximum absolute peak, and writes the top files to top_peak_files.txt.

• Set the input folder and output file in .env if needed.
• Run it on the folder.
• Check the saved peak report.

Screens noisy takes and identifies recordings with elevated noise floor for later review or removal.

• Run it against the target folder of recordings.
• Review the saved output for flagged noisy files.
• Use it before final dataset export.