Speech Indexer

---

=> 日本語

 

Speech Indexer is a Windows GUI program that simplifies the task of developing and exploring speech/video corpora with the aid of OpenAI Whisper, an automatic speech recognition system. Speech Indexer utilizes 'Whisper.cpp' created by Georgi Gerganov, which I've compiled using MinGW-w64 for it to run on the Windows platform. To learn more about Whisper and Whisper.cpp, please visit theirt respective websites: Whisper and Whisper.cpp.

 

System Requirements

OS：Microsoft Windows 10, 11

.NET Framework： 4.7.2 or later (If not installed, please download the runtime from here)

Screen resolution: 960 * 640 or higher

Windows Media Player

 

Setup Instructions

The transcription performance depends on which model you use. The baseline model file 'ggml-base.bin' is bundled with the program, so it's ready to use once it's downloaded. Other model files can be obtained from 'Hugging Face'. Bear in mind, the larger the model file you use, the higher the recognition accuracy, but the processing time will increase accordingly. Please store any model files you download on the 'models' subfolder of the 'Whisper.cpp' folder.

 

ggml-tiny.bin (77.7MB)

ggml-base.bin (148MB) ※included

ggml-small.bin (488MB)

ggml-medium.bin (1.53GB)

ggml-large-v1.bin (3.09GB)

 

Uninstallation Instructions

Simply delete the 'SpeechIndexer' folder completely. Rest assured that the program makes no modifications to your system's registry.

 

Supported Media Formats

Audio Formats: WAV, MP3, M4A

Video Formats: MP4, MOV

 

Third-party Software

Speech Indexer makes use of the following software:

Whisper:

  Copyright (c) 2022 OpenAI, Released under the MIT License,

  see https://github.com/openai/whisper/blob/main/LICENSE

 

Whisper.cpp:

  Copyright (c) 2023 Georgi Gerganov, Released under the MIT License,

  see https://github.com/ggerganov/whisper.cpp/blob/master/LICENSE

  *Speech Indexer incorporates Whisper.cpp, with an additional feature I've implemented.

 

NAudio:

  Copyright (c) 2020 Mark Heath, Released under the MIT License,

  see https://github.com/naudio/NAudio/blob/master/license.txt

 

CsvHelper:

  Copyright (c) 2009-2022 Josh Close, Released under the Apache v2.0 license,

  see https://github.com/JoshClose/CsvHelper/blob/master/LICENSE.txt

 

Citation Guidelines

If citing, please use the following format:

・Goto, K. (2023). Speech Indexer (Version 0.1) [Computer Software]. Retrieved from: https://www.setsunan.ac.jp/~corpus/SpeechIndexer_en.htm

 

Usage Instructions

Speech Indexer offers two man modes: "Corpus Builder" and "Keyword Search." Users can switch between these modes by selecting the tabs in the tool bar.

 

How to Transcribe Files

This section shows you how to use the "Corpus Builder" mode for transcription.

 

 

(1), (2)

Clicking on the "Files" button in the file menu (shown as (1) in the image above) will open a context menu with the following options.

Add Files:This allows you to add file(s) to the "Selected File" list (shown as (2) in the image above).

Remove Files: This allows you to remove selected file(s) from the "Selected File" list.

Clear All Files: This allows you to clear all files from the "Selected File" list.

 

(3): Transcription Settings

Model: Choose one out of the five models listed above to use for transcription.

Language: Choose the language you want to transcribe: English, Japanese, or Others.

Output Format: Choose among "TSV," "CSV," or "TXT" for your output format. Note that Speech Indexer does not support keyword searches on plain TXT files.

Segment length (optional): Specify the length of the generated text segments.

Prompt (optional): Keep the filler words in your transcript. You can specify which fillers to include by editing "Filler_ex.txt" in the "Prompt" folder. See OpenAI for more information on prompts.

 

(4), (5)

Transcribe: Click on the "Transcribe" button (shown as (4) in the image above) to start the transcription. You can monitor the ongoing task's progress in the "Console" area (shown as (5) in the image above).

Cancel: Stop the ongoing transcription process.

 

How to Search Transcripts for Keywords

This section shows you how to use the "Keyword Search" mode.

 

 

(6), (8)

Clicking on the "Files" button in the file menu (shown as (6) in the image above) will open a context menu with the following options.

Add Files: This allows you to add file(s) for keyword search to the "Selected File" list (shown as (8) in the image above).

Remove Files: This allows you to remove selected files from the "Selected File" list.

Clear All Files: This allows you to clear all files from the "Selected File" list.

 

(7): Enter a search string, which can be either a single word or a regular expression.

 

(9): You can see the preview of the file in the "Preview" area (shown as (9) in the image above) by double-clicking on it in the "Selected File" list.

 

(10): The table (shown as (10) in the image above) shows search results with these headings: "File Path," "Start," "End," and "Text". By double-clicking on any of these lines, the corresponding audio/video file will start playing in a separate window.

 

How to Navigate through Media Files

This function allows you to find how a word or phrase you search for is spoken in specific contexts.

 

 

(11): Windows Media Player is displayed in "Scene Viewer" (shown as (11) in the image above) to play a media file.

 

(12)

Repeat: If you want the current segment to loop, tick the checkbox.

Edit Mode: Ticking the checkbox will enable you to edit the displayed file.

 

(13):The table (shown as (13) in the image above) shows transcription details: "File," "Start," "End," and "Text" for the current and several subsequent segments. Double-clicking on any of these segments to play the corresponding scene.

 

 

 

Update History

Version 0.22: Dec. 2023

Version 0.21: Dec. 2023

Version 0.2: Oct. 2023

Version 0.11: Sep. 2023

Version 0.1: Sep. 2023

 

Copyright (C) GOTO Kazuaki, 2004 - 2023 All rights reserved