Update README.md

GUSSIED IT UP BOI!!

README.md

@@ -4,125 +4,72 @@ tags:
 - torch
 - cuda
 - identifcation
-- forensics
-license: cc-by-nc-nd-4.0
+license: cc-by-nc-4.0
 datasets:
 - pyannote/speaker-diarization
 language:
 - en
-pretty_name: Speaker Diarization Toolkit
+pretty_name: Speaker Identification Toolkit
 ---
-# Speaker Identification Toolkit (Speaker Diarization)
+# Speaker Identification Toolkit
 
-This repository contains Python scripts for processing audio and video files, focusing on speaker diarization, speaker identification, audio extraction, and dataset creation. These tools leverage `ffmpeg`, `pyannote.audio`, and other Python libraries for efficient and accurate processing.
+## Overview
 
-## Scripts Overview
+This toolkit provides a comprehensive suite of scripts for audio processing and speaker identification tasks. It includes features like video-to-audio conversion, speaker diarization, speaker isolation, and audio trimming.
 
-### 1. **dataset-creation.py**
+## Scripts
 
-- **Description**: Extracts English audio tracks from video files and converts them to mono WAV format for further processing.
-- **Key Features**:
-  - Processes video files in `.mp4` and `.mkv` formats.
-  - Outputs audio in WAV format (16kHz, mono).
-- **Configuration**:
-  - Edit the paths for `INPUT_DIR` (video files directory) and `OUTPUT_DIR` (audio output directory) directly in the script before running.
-- **Dependencies**: `rich`, `ffmpeg`
+### 1. Dataset Creation (Video to Audio Extraction)
 
----
-
-### 2. **diarize-dataset.py**
-
-- **Description**: Processes WAV files to generate JSON files containing speaker diarization data.
-- **Key Features**:
-  - Uses Hugging Face's `pyannote.audio` for speaker diarization.
-  - Outputs JSON files with speaker labels and timestamps.
-- **Configuration**:
-  - Edit the paths for `INPUT_DIR` (WAV files directory) and `OUTPUT_DIR` (JSON output directory) directly in the script before running.
-  - Set the `HF_TOKEN` variable to your Hugging Face API token.
-- **Dependencies**: `pyannote.audio`, `rich`
-
----
-
-### 3. **identify-speaker.py**
-
-- **Description**: Maps a target speaker from JSON diarization files by listening to audio segments interactively.
-- **Key Features**:
-  - Plays audio segments for each speaker and allows the user to select the target speaker.
-  - Maintains a mapping of files to target speakers in a CSV file.
-- **Configuration**:
-  - Edit the paths for `JSON_DIR` (JSON files directory), `AUDIO_DIR` (WAV files directory), and `MAPPING_FILE` (CSV file for speaker mappings) directly in the script before running.
-- **Dependencies**: `playsound`, `rich`, `pandas`, `ffmpeg`
-- **Recommendations**:
-  - Listen to multiple segments before confirming that it's your targeted speaker; it can sometimes have 1 second of your speaker followed by 9 seconds of your non-targeted speaker, and will throw off your dataset.
-  - If you make a mistake, stop processing, and remove the bottom most entry from the speaker\_mappings.csv file, and then re-run the script.
-
----
-
-### 4. **isolate-trim.py**
-
-- **Description**: Extracts and trims audio segments of the target speaker, splitting them into smaller clips if necessary.
-- **Key Features**:
-  - Extracts segments for a target speaker based on the mapping CSV.
-  - Crops audio into chunks of a specified length and trims silence from each chunk.
-- **Configuration**:
-  - Edit the paths for `JSON_DIR` (JSON files directory), `AUDIO_DIR` (WAV files directory), `MAPPING_FILE` (CSV file for speaker mappings), and `OUTPUT_DIR` (output directory for processed files) directly in the script before running.
-- **Dependencies**: `rich`, `pandas`, `ffmpeg`
-
----
-
-## Setup
-
-### Install Dependencies
-
-Ensure you have the following installed:
-
-- Python 3.8+
-- Required Python libraries:
-  ```bash
-  pip install rich pandas pyannote.audio playsound
-  ```
-- `ffmpeg`:
-  Install `ffmpeg` via your system's package manager or [official site](https://ffmpeg.org/).
-
-### Configuration
+- **Functionality:** Extracts audio tracks from video files.
+- **Default Directories:**
+  - Input: `videos`
+  - Output: `wavs`
+- **Behavior:**
+  - Auto-creates `videos` and `wavs` directories in the script folder.
+  - Pauses execution to allow the user to populate the `videos` directory before processing.
+  - Processes all `.mkv` and `.mp4` files in the `videos` directory.
 
-Each script requires paths to be manually edited in the file before execution. Below is the default directory structure:
+### 2. Speaker Diarization
 
-- **Video Input Directory**: `base-folder/videos`
-- **WAV Output Directory**: `base-folder/wavs`
-- **JSON Output Directory**: `base-folder/jsons`
-- **Speaker Mapping File**: `base-folder/scripts/speaker_mapping.csv`
-- **Processed Speaker Output Directory**: `base-folder/targeted`
+- **Functionality:** Processes audio files and generates speaker diarization metadata.
+- **Default Directories:**
+  - Input: `wavs`
+  - Output: `jsons`
+- **Behavior:**
+  - Auto-creates `wavs` and `jsons` directories in the script folder.
+  - Includes a prompt for the Hugging Face token with validation.
+  - Pauses execution to allow the user to populate the `wavs` directory.
+  - Generates diarization metadata in `.json` format for all `.wav` files in the `wavs` directory.
 
-Additionally:
+### 3. Speaker Isolation and Trimming
 
-- For `diarize-dataset.py`, set the Hugging Face API token in the script:
-  ```python
-  HF_TOKEN = "your_hugging_face_token"
-  ```
+- **Functionality:** Isolates and trims audio segments based on speaker diarization metadata.
+- **Default Directories:**
+  - Input JSON: `jsons`
+  - Input WAV: `wavs`
+  - Output: `targeted`
+- **Behavior:**
+  - Auto-creates `jsons`, `wavs`, and `targeted` directories in the script folder.
+  - Pauses execution to allow the user to populate the `jsons` and `wavs` directories.
+  - Processes each `.json` file and its corresponding `.wav` file to extract and trim speaker-specific segments.
+  - Outputs trimmed audio clips in the `targeted` directory.
 
----
-
-## Workflow
-
-1. **Extract Audio from Videos**:
-   Use `dataset-creation.py` to extract English audio tracks from video files.
+## Common Features
 
-2. **Generate Speaker Diarization Data**:
-   Use `diarize-dataset.py` to process the extracted WAV files and generate JSON diarization files.
+- All scripts auto-create necessary directories in the same folder as the script.
+- Scripts pause execution at key points to allow the user to populate input directories.
+- Rich console output for enhanced user interaction and error handling.
 
-3. **Identify Target Speaker**:
-   Execute `identify-speaker.py` to map each JSON file to a target speaker interactively.
+## Hugging Face Token
 
-4. **Isolate and Trim Audio**:
-   Run `isolate-trim.py` to extract and process audio segments for the target speaker.
-
----
+- Required for the Speaker Diarization script.
+- Token validation is performed before processing.
+- Update your token in the script or provide it at runtime when prompted.
 
 ## Notes
 
-- Ensure all required directories exist and paths are correctly set before running the scripts.
-- High-quality audio improves the accuracy of diarization and speaker identification.
-- The scripts are designed for processing local files and do not include automated cloud integration or downloads.
-
----
+- Ensure all input files are placed in the correct directories before running the scripts.
+- Output files will be saved in the corresponding output directories, which are automatically created if they do not exist.
+- When processing multi-speakers in multi-episode files, like a TV series spanning multiple seasons, it is best practice to listen to a few segments of the speaker before confirming.
+- If you make a mistake, stop the script and remove the bottom most line in your mappings CSV; the script will re-process it on next run.