WhisperStream User Guide

Getting Started

WhisperStream is a Windows application that provides powerful audio transcription and text-to-speech capabilities. When you first launch the application, you'll see the main menu with options for different operations.

Main Menu

The main menu is your starting point for all operations in WhisperStream. It provides quick access to all major features:

• Single File Speech-to-Text: Transcribe a single audio file to text
• Batch Speech-to-Text: Process multiple audio files in a batch operation
• Single File Text-to-Speech: Convert text to speech for a single file
• Batch Text-to-Speech: Convert multiple text files to speech
• Exit: Close the application

Simply click on any button to open the corresponding form and begin your work.

Single File Speech-to-Text

This screen allows you to transcribe a single audio file into text using Whisper AI models.

Using Single File Speech-to-Text

1. Select an Audio File: Click the "Browse" button to select an audio file. Supported formats include WAV, MP3, M4A, OGG, FLAC, and MP4.
2. Choose a Model: Select a Whisper model from the dropdown:
   • Tiny: Fastest processing, least accurate
   • Base: Good balance of speed and accuracy (recommended for most users)
   • Small: Better accuracy, moderate processing time
   • Medium: High accuracy, slower processing
   • Large: Best accuracy, slowest processing
3. Transcribe: Click the "Transcribe" button to start the transcription process. The status bar will show progress updates.
4. Review Results: Once complete, the transcribed text will appear in the output area.
5. Save Transcript: Click "Save" to save your transcript in various formats:
   • Plain Text (.txt)
   • SubRip Subtitles (.srt)
   • WebVTT Subtitles (.vtt)
   • CSV Data (.csv)
   • Timestamped Text (.txt)

Menu Options

Batch Speech-to-Text

The Batch Speech-to-Text form allows you to process multiple audio files simultaneously, making it ideal for transcribing large collections of audio files.

Using Batch Speech-to-Text

1. Add Files: Click "Add Files" to select multiple audio files at once. All selected files will be added to the job queue.
2. Configure Parallel Processing: Set the maximum number of parallel jobs (1-8). Higher values process faster but use more system resources.
3. Start Processing: Click "Start Processing" to begin transcribing all queued files. The status bar shows overall progress.
4. Monitor Progress: The job grid displays:
   • File name
   • Status (Queued, Processing, Completed, Failed, Cancelled)
   • Progress percentage
   • File size
   • Status message
   • Duration
   • Error messages (if any)
   • Transcript text preview
5. Manage Jobs:
   • Cancel Selected: Cancel selected jobs
   • Retry Selected: Retry failed or cancelled jobs
   • Remove Selected: Remove jobs from the queue
   • Remove Completed: Clear completed jobs from the list
   • Clear All: Remove all jobs (when not processing)
6. Export Results: Click "Export Results" to save all completed transcripts to a folder. You can choose to remove filler words during export.

Job Grid Features

Single File Text-to-Speech

Convert text into natural-sounding speech using voice cloning technology. This feature uses Chatterbox-TTS for high-quality voice synthesis.

Using Single File Text-to-Speech

1. Enter Text: Type or paste the text you want to convert to speech in the input area.
2. Select Voice Prompt (Optional):
   • Click "Browse" to select a voice sample file (WAV format)
   • The voice sample will be used to clone the voice characteristics
   • You can also enter a text description of the desired voice
3. Choose Output Format: Select the audio format for the output file (WAV, MP3, etc.)
4. Generate Audio: Click "Generate" to start the TTS process. A save dialog will appear to choose where to save the output file.
5. Play Audio: Once generation is complete, click "Play Audio" to preview the generated speech. Click again to stop playback.

Additional Features

Single File Text-to-Speech interface

Batch Text-to-Speech

Process multiple text files or text entries simultaneously, converting them all to speech audio files. This is ideal for creating large collections of audio content.

Setting Up Batch TTS

1. Select Output Directory: Click "Browse" next to the output directory field to choose where generated audio files will be saved.
2. Choose Output Format: Select the audio format (WAV, MP3, etc.) for all generated files.
3. Select Default Voice: Choose a default voice from the dropdown. This voice will be used for all jobs unless overridden.
4. Open Voices Folder: Click "Open Voices Folder" to access the directory where voice samples are stored. You can organize voices in subfolders.

Adding Jobs

Hello world|subfolder/voice1
This is another line|voice2
This line uses the default voice

Job Management

The job grid displays:

• Text: Preview of the text content
• Voice: Voice file name being used
• Status: Current processing status
• Started: When processing began
• Completed: When processing finished
• Output File: Path to the generated audio file

Processing Jobs

1. Start Processing: Click "Start" to begin processing all queued jobs. Jobs are processed sequentially.
2. Monitor Progress: Watch the progress bar and status messages to track processing.
3. Cancel Processing: Click "Cancel" to stop processing. Current job will finish, but queued jobs will remain.

Managing Results

Batch Text-to-Speech interface

Media Player

The Media Player form provides synchronized audio playback with transcript editing, making it perfect for reviewing and editing transcriptions while listening to the original audio.

Opening the Media Player

You can open the Media Player from:

• The Tools menu in the Single File Speech-to-Text form
• Directly from the main menu (if available)

Using the Media Player

1. Open Audio File: Use File → Open Audio File or the toolbar button to load an audio file (WAV, MP3, M4A, FLAC, OGG, AAC).
2. Load Transcript (Optional): Use File → Load Transcript to load an existing transcript file (JSON or TXT format).
3. Playback Controls: The waveform display shows the audio waveform. Click on the waveform to seek to specific positions.
4. Edit Transcript: The transcript editor allows you to edit the text while the audio plays. Changes are synchronized with playback position.
5. Save Transcript: Use File → Save Transcript to save your edited transcript with timing information.
6. Export Transcript: Use File → Export Transcript to export as plain text without timing data.

Menu Options

Features

Settings

The Settings form allows you to configure all aspects of WhisperStream. Access it from the Tools menu in any form.

General Tab

Transcription Tab

Audio Tab

UI Tab

Privacy Tab

Advanced Tab

Saving Settings

• OK: Save all settings and close the dialog
• Cancel: Discard changes and close without saving
• Apply: Apply settings without closing the dialog
• Reset: Reset all settings to their default values

Diagnostics

The Diagnostics form helps identify and resolve setup issues with required components like Python, Whisper, and FFmpeg.

Understanding Diagnostics

When you first launch WhisperStream or if there are setup issues, the Diagnostics form will display:

• A list of any missing or improperly configured components
• Recommended installation steps
• Buttons to automatically install missing components

Available Actions

Installation Progress

When installing components, a progress window displays:

• Real-time command-line output
• Progress percentage
• Status messages
• Error information (if any)

You can save the installation log for troubleshooting by clicking "Save Log" in the progress window.

Tips & Tricks

Getting the Best Transcription Results

• Audio Quality: Use clear audio with minimal background noise for best results
• Model Selection: Start with "base" model for most use cases. Use "large" only when accuracy is critical and processing time is not a concern
• Language Settings: Specify the language if you know it, rather than using auto-detect, for better accuracy
• Speaker Diarization: Enable this feature when transcribing conversations with multiple speakers

Optimizing Performance

• Batch Processing: Adjust the parallel jobs setting based on your system's capabilities. More parallel jobs = faster processing but higher CPU/memory usage
• GPU Acceleration: If you have a compatible GPU, enable GPU acceleration in Advanced settings for significantly faster transcription
• Memory Limits: Adjust memory limits in Advanced settings if you encounter out-of-memory errors

Text-to-Speech Best Practices

• Voice Samples: Use 3-10 second voice samples with clear, natural speech
• Text Formatting: Use proper punctuation and capitalization for better speech synthesis
• Batch Processing: For large projects, use Batch TTS to process multiple files efficiently
• Voice Organization: Organize voice files in subfolders within the voices directory for easy management

Workflow Tips

• Save Frequently: Always save your transcripts after transcription to avoid losing work
• Use Media Player: Use the Media Player to review and edit transcripts while listening to the original audio
• Export Formats: Export in multiple formats (SRT for video editing, TXT for text processing, etc.)
• Batch Operations: Use batch processing for large collections of files to save time

Troubleshooting

Conclusion

This guide covers all the major features and screens in WhisperStream. The application is designed to be intuitive and user-friendly, with tooltips and help text available throughout the interface.

For the best experience:

• Take time to explore the Settings to customize WhisperStream to your needs
• Use batch processing for efficiency when working with multiple files
• Take advantage of the Media Player for reviewing and editing transcripts
• Keep your components (Python, Whisper, FFmpeg) up to date

Happy transcribing! We hope WhisperStream helps you efficiently convert between speech and text for all your projects.