SpeechPulse Updates

4/20/2025 - Version 10.7.1 (Windows)

• System audio mode now includes two collapsible sections (Mic and System) to display the full transcription for better readability.

4/18/2025 - Version 10.7.0 (Windows)

• SpeechPulse now supports manual punctuation for the German language (in addition to English).
• Supported manual punctuation commands are listed in the voice command guide (Settings->Voice command guide).
• Manual punctuation will be available for other languages in the future.

4/16/2025 - Version 10.6.3 (Windows)

• Fixed an issue that caused SpeechPulse to crash for some specific audio segments under certain (rare) conditions.

4/14/2025 - Version 10.6.2 (Windows)

• Fixed an issue that caused SpeechPulse to randomly crash with some specific audio files and speech segments.
• A new option to reduce the CPU usage when running on an NVIDIA GPU.

4/11/2025 - Version 10.6.0 (macOS)

• System audio mode now supports text selection, copying, and finding operations while the transcription is running.
• System audio mode now supports pause and resume operations.
• SpeechPulse now remembers the last used mode (from Live mode, File mode, and System audio mode) and automatically loads into that mode.
• System audio mode now skips recording to WAV files while pausing.

4/11/2025 - Version 10.6.1 (Windows)

• System audio mode now skips recording to WAV files while pausing.

4/10/2025 - Version 10.6.0 (Windows)

• System audio mode now supports text selection, copying, and finding operations while the transcription is running.
• System audio mode now supports pause and resume operations.
• SpeechPulse now remembers the last used mode (from Live mode, File mode, and System audio mode) and automatically loads into that mode.
• A new option to disable the "SpeechPulse Minimized" notification.

4/8/2025 - Version 10.5.4 (macOS)

• UI improvements.

4/8/2025 - Version 10.5.4 (Windows)

• Reduced CPU usage when running on an NVIDIA GPU.

4/6/2025 - Version 10.5.3 (Windows)

• UI improvements.

4/3/2025 - Version 10.5.2 (macOS)

• You can now back up all of your SpeechPulse settings and data, including app settings, training profiles, training data, trained models, dictation history, and saved transcriptions/diarizations, using the new backup option.
• The backup operation saves all settings and data to a single archive file on your local disk drive.
• You can later restore SpeechPulse (on the same computer or on a different one) using the new restore option.
• SpeechPulse doesn't support backup/restore between Windows and macOS (both computers should run Windows or both should run macOS).
• A new option to increase the paste mode clipboard wait time to prevent any text insertion issues.

4/3/2025 - Version 10.5.2 (Windows)

• Added a disk synchronization step before the backup operation. This will avoid any settings mismatch in a rare case of OS caching.

4/1/2025 - Version 10.5.1 (Windows)

• Added support for the NVIDIA GPUs with Compute Capability 5.x or lower.

3/27/2025 - Version 10.5.0 (Windows)

• You can now back up all of your SpeechPulse settings and data, including app settings, training profiles, training data, trained models, dictation history, and saved transcriptions/diarizations, using the new backup option.
• The backup operation saves all settings and data to a single archive file on your local disk drive.
• You can later restore SpeechPulse (on the same computer or on a different one) using the new restore option.
• SpeechPulse doesn't support backup/restore between Windows and macOS (both computers should run Windows or both should run macOS).

3/23/2025 - Version 10.4.2 (Windows)

• Added support for the new RTX 50 series NVIDIA GPUs.
• GPU execution requires newer CUDA libraries, which you can download using the built-in library downloader.
• A new option to enable a faster and higher accuracy computation mode on NVIDIA GPUs with Compute Capability 7.0 or higher and 6 GB or larger VRAM.
• A new option to increase the paste mode clipboard wait time to prevent any text insertion issues.

3/16/2025 - Version 10.4.0 (macOS)

• Now supports training for the French, Italian, Spanish, and Dutch languages (in addition to English and German).
• Added voice commands for non-English languages.
• For example, now you can say "Neue Zeile" to add a new line and "Neuer Absatz" to add a new paragraph while dictating in German.
• Supported voice commands are listed for each language in the voice command guide (Settings->Voice command guide).
• Several improvements to prevent missing punctuation marks in non-English languages.
• Fixed an issue that caused the Text inserter to insert multiple times in the real-time processing mode.
• Fixed an issue that caused the "New line" and "New paragraph" commands to fail when dictated after a pause.
• Misc. improvements.

3/11/2025 - Version 10.4.0 (Windows)

• Added voice commands for more languages.
• Fixed an issue that caused SpeechPulse to fail with conflicting CUDA libraries.
• Misc. improvements.

3/6/2025 - Version 10.3.0 (Windows)

• Now supports voice commands for the German, French, Italian, Spanish, and Dutch languages (in addition to English).
• For example, now you can say "Neue Zeile" to add a new line and "Neuer Absatz" to add a new paragraph while dictating in German.
• Supported voice commands are listed for each language in the voice command guide (Settings->Voice command guide).
• Voice commands will be available for other languages in the future.

3/5/2025 - Version 10.2.0 (Windows)

• Now supports training for the French, Italian, Spanish, and Dutch languages (in addition to English and German).
• Several improvements to prevent missing punctuation marks in non-English languages.
• Fixed an issue that caused the Text inserter to insert multiple times in the real-time processing mode.
• Fixed an issue that caused the "New line" and "New paragraph" commands to fail when dictated after a pause.
• Misc. improvements.

2/27/2025 - Version 10.1.0 (macOS)

• Now supports training for the German language.
• Misc. improvements.

2/25/2025 - Version 10.1.0 (Windows)

• Now supports training for the German language.
• Misc. improvements.

2/22/2025 - Version 10.0.6 (macOS)

* Training on the SpeechPulse macOS version is currently running on the CPU. We will try to add GPU training in the future for faster training.

• A new feature to support training new words to SpeechPulse.

2/18/2025 - Version 10.0.6 (Windows)

* The first training run in this new version will take slightly longer, depending on the number of words in the speech profile.

• Several improvements for significantly better training accuracy.
• Fixed an issue that caused training to get stuck at a suboptimal solution.

2/16/2025 - Version 10.0.5 (Windows)

• Fixed an issue that caused the UI to slow down.
• Fixed an issue that caused poor training accuracy.

2/15/2025 - Version 10.0.4 (Windows)

• Reduced the disk space required for training data.
• Misc. improvements for training.

2/12/2025 - Version 10.0.3 (Windows)

• Several improvements to make the training faster and more accurate.

2/11/2025 - Version 10.0.2 (Windows)

• Several improvements for better training accuracy.

2/11/2025 - Version 10.0.1 (Windows)

• Several improvements to the training feature.

2/10/2025 - Version 10.0.0 (Windows)

• A new feature to support training new words to SpeechPulse.

1/19/2025 - Version 9.4.1 (Windows and macOS)

• Transcription and diarization editors now export the edited subtitle files.
• Misc. improvements.

1/8/2025 - Version 9.4.0 (macOS)

• A new dynamic AI template to read the AI instructions from your dictation.
• For example, you can say, "Correct grammar and spelling in the last dictated text," to correct any grammar/spelling issues in your last dictated text.
• Similarly, you can say, "Correct grammar and spelling in the clipboard text," to correct any grammar/spelling issues in the text you have already copied to the clipboard (e.g., using CTRL+C).

1/7/2025 - Version 9.4.0 (Windows)

• A new dynamic AI template to read the AI instructions from your dictation.
• For example, you can say, "Correct grammar and spelling in the last dictated text," to correct any grammar/spelling issues in your last dictated text.
• Similarly, you can say, "Correct grammar and spelling in the clipboard text," to correct any grammar/spelling issues in the text you have already copied to the clipboard (e.g., using CTRL+C).

1/5/2025 - Version 9.3.2 (macOS)

• UI improvements.
• Several performance improvements for the transcription window.
• Fixed an issue that causes higher memory usage in the transcription and diarization windows.
• Fixed an issue that causes higher memory usage in the system audio mode.

1/5/2025 - Version 9.3.2 (Windows)

• Several performance improvements for the transcription window.
• Fixed an issue that causes higher memory usage in the system audio mode.

1/2/2025 - Version 9.3.1 (Windows)

• Fixed an issue that can cause missing speech segments around silent parts of the audio (Live mode, File mode).
• Fixed an issue that causes higher memory usage in the transcription and diarization windows.
• Misc. improvements.

12/30/2024 - Version 9.3.0 (Windows)

• File mode now supports adding new files, removing pending files, and reordering the file list using drag and drop while the transcription is running.
• UI improvements.

12/27/2024 - Version 9.2.1 (Windows)

• UI improvements (System audio mode).
• A new hotkey to enable real-time processing.
• Misc. improvements.

12/27/2024 - Version 9.2.1 (macOS)

• UI improvements (System audio mode).
• A new hotkey to enable real-time processing.
• Misc. improvements.

12/25/2024 - Version 9.2.0 (macOS)

• A new speaker diarization implementation for the file mode (in addition to the original implementation).
• The new diarization implementation supports setting active speakers per file for better accuracy. It is also significantly faster than the original implementation.
• Real-time processing is now more accurate and responsive (live mode and system audio mode).
• Fixed an issue of missing speech segments in the real-time live mode and system audio mode.
• Misc. improvements.

12/20/2024 - Version 9.2.0 (Windows)

• A new speaker diarization implementation for the File mode (in addition to the original implementation).
• The new diarization implementation supports setting active speakers per file for better accuracy.
• The new diarization implementation is significantly faster than the original implementation.
• You can switch between the two diarization implementations using the File mode UI.

12/18/2024 - Version 9.1.2 (macOS)

• Automatic speaker diarization in the system audio mode can segment the transcription for each individual speaker.
• Several improvements to the automatic speaker diarization (File mode and system audio mode).
• Fixed an issue that caused the push-to-talk auto mic off function to fail.
• UI improvements.

12/18/2024 - Version 9.1.2 (Windows)

• UI improvements.

12/17/2024 - Version 9.1.1 (Windows)

• Several improvements to the automatic speaker diarization (File mode and system audio mode).
• UI improvements.

12/16/2024 - Version 9.1.0 (Windows)

• Several improvements to the automatic speaker diarization (File mode and system audio mode).
• Fixed an issue that caused slightly lower accuracy and higher delay in the live dictation mode.

12/12/2024 - Version 9.0.1 (Windows)

• Fixed an issue that caused SpeechPulse to crash with some non-English languages.

12/09/2024 - Version 9.0.0 (Windows)

• Automatic speaker diarization in the system audio mode can segment the transcription for each individual speaker.

12/03/2024 - Version 8.3.0 (macOS)

• Diarization and transcription editors now use the left mouse click to set the audio player position.
• Diarization and transcription editors now support play/pause using the Tab key.
• The diarization editor now uses the Enter key to split segments.
• Pressing the Backspace key at the beginning of a diarization segment will combine it with the previous segment.
• UI improvements.

12/03/2024 - Version 8.3.0 (Windows)

• Diarization and transcription editors now use the left mouse click to set the audio player position.
• Diarization and transcription editors now support play/pause using the Tab key.
• The diarization editor now uses the Enter key to split segments.
• Pressing the Backspace key at the beginning of a diarization segment will combine it with the previous segment.
• UI improvements.

11/30/2024 - Version 8.2.3 (Windows)

• A new option to select a non-default output device as the system audio source.

11/30/2024 - Version 8.2.2 (macOS)

• Several improvements/fixes for the system audio mode.

11/30/2024 - Version 8.2.2 (Windows)

• Several improvements/fixes for the system audio mode.

11/29/2024 - Version 8.2.1 (macOS)

• Diarization and transcription editors now support the "Find and Replace" function.
• Now supports manually setting speaker counts for diarization.
• The Minimize/Maximize hotkey is now remapped to the "Minimize to tray/Maximize from tray" functionality.
• New options to remove silent/noisy audio segments and reduce text hallucinations in the system audio mode.
• Fixed an issue that caused incorrect spacing in the system audio mode.

11/28/2024 - Version 8.2.1 (Windows)

• Fixed an issue that caused incorrect spacing in the system audio mode.

11/28/2024 - Version 8.2.0 (Windows)

• Diarization and transcription editors now support the "Find and Replace" function.
• Now supports manually setting speaker counts for diarization.
• The Minimize/Maximize hotkey is now remapped to the "Minimize to tray/Maximize from tray" functionality.
• New options to remove silent/noisy audio segments and reduce text hallucinations in the system audio mode.

11/24/2024 - Version 8.1.4 (macOS)

• A new option to enable high-accuracy timestamps for the system audio mode.
• A new option to enable high-accuracy timestamps for the Real-time Live Mode.

11/23/2024 - Version 8.1.3 (macOS)

• A new option to combine microphone and system transcriptions into a single editor in the system audio mode.
• A new option to enable tagging/highlighting microphone and system transcriptions in the combined system audio mode.
• A new option to set a silence timeout for the system audio mode. SpeechPulse will consider a speech segment is complete after this duration of silence.
• A new option to automatically restart the system transcription in case the system audio stream is stopped.
• Misc. improvements.

11/20/2024 - Version 8.1.3 (Windows)

• Several improvements for the system audio mode.
• Misc. improvements.

11/19/2024 - Version 8.1.2 (Windows)

• A new option to enable tagging/highlighting microphone and system transcriptions in the combined system audio mode.
• A new option to set a silence timeout for the system audio mode. SpeechPulse will consider a speech segment is complete after this duration of silence.
• Disabled the mouse wheel scrolling of dropdown controls to prevent accidental selection.
• Misc. improvements.

11/16/2024 - Version 8.1.1 (Windows)

• A new option to combine microphone and system transcriptions into a single editor in the system audio mode.
• Several performance improvements (reduced CPU and RAM usage).

11/13/2024 - Version 8.1.0 (Windows)

• Timestamps in system audio mode are now more accurate.
• System audio mode now displays a processing message if any audio remains to be transcribed.

11/10/2024 - Version 8.0.8 (macOS)

• Several performance improvements for transcription and diarization UIs.

11/10/2024 - Version 8.0.8 (Windows)

• Several performance improvements for transcription and diarization UIs.
• Fixed an issue that caused paragraph segmentation to fail.

11/09/2024 - Version 8.0.7 (macOS)

• System audio mode supports real-time transcription of mic and system audio to an internal editor (no mouse focus required).
• Record mic and system audio to WAV files.
• Automatic paragraph segmentation and improved sentence segmentation.
• The diarization edit speaker names window now suggests existing speaker names when you type.
• Fixed an issue that caused diarization (via API) to fail with long audio segments.
• Supports translation to English via Whisper APIs (in addition to via offline models).
• Added support for some missing audio file formats, including Opus.
• A new hotkey to add text replacements (mappings) without opening the settings window.
• Fixed an issue that caused high memory usage on transcription and diarization windows.
• Transcription and diarization windows are now more responsive (reduced CPU and RAM usage).
• Misc. improvements/fixes.

11/08/2024 - Version 8.0.7 (Windows)

• Fixed an issue that caused high memory usage on transcription and diarization windows.
• Transcription and diarization windows are now more responsive (reduced CPU and RAM usage).

11/03/2024 - Version 8.0.6 (Windows)

• Several improvements to prevent missing punctuation marks in Live Mode, File Mode, and system audio mode.

11/02/2024 - Version 8.0.5 (Windows)

• Fixed an issue that caused SpeechPulse to reset model/device settings under certain conditions.

10/30/2024 - Version 8.0.4 (Windows)

• Updated the Multi (turbo) model with a newer, more accurate version. You can delete the older model and download the new one using the built-in model downloader.
• Fixed an issue that caused SpeechPulse to hang in the system audio mode on some PCs/laptops.
• Paragraph segmentation is now skippable (to reduce transcription duration for lengthy audio files).
• Several improvements to prevent missing punctuation marks in lengthy audio files (English language).

10/29/2024 - Version 8.0.3 (Windows)

• Reduce text hallucinations in the Live Mode and File Mode.
• Misc. improvements.

10/28/2024 - Version 8.0.2 (Windows)

• Reduce text hallucinations in the system audio mode.
• System audio mode now supports timestamps.

10/26/2024 - Version 8.0.1 (Windows)

• Several performance/accuracy improvements for the system audio mode.

10/25/2024 - Version 8.0.0 (Windows)

• System audio mode supports real-time transcription of mic and system audio to an internal editor (no mouse focus required).
• Record mic and system audio to WAV files.
• Automatic paragraph segmentation and improved sentence segmentation.
• The diarization edit speaker names window now suggests existing speaker names when you type.
• Fixed an issue that caused diarization (via API) to fail with long audio segments.
• Supports translation to English via Whisper APIs (in addition to via offline models).
• Added support for some missing audio file formats, including Opus.
• A new hotkey to add text replacements (mappings) without opening the settings window.
• Misc. improvements/fixes.

10/14/2024 - Version 7.1.0 (macOS)

• Now supports real-time transcription in live mode (Experimental - Currently has a high delay when inserting/editing the dictated text due to input buffering on macOS).
• A new option to edit/delete the saved diarization speakers.
• Fixed an issue that caused SpeechPulse to crash when the transcription contains invalid characters.

10/5/2024 - Version 7.1.0 (Windows)

• SpeechPulse (Windows) now supports the Whisper V3 (turbo) model. You can download it using the built-in model downloader.
• This model has comparable accuracy to the Whisper (large) model and runs significantly faster.
• Real-time processing mode now only updates the changed portion of the dictation, making dictation less distracting.

10/4/2024 - Version 6.3.3 (macOS)

• SpeechPulse (macOS) now comes with the Whisper V3 (turbo) model as the default model. This model has comparable accuracy to the Whisper (large) model and runs significantly faster.

10/1/2024 - Version 7.0.1 (Windows)

• Several performance improvements for the real-time processing.

9/30/2024 - Version 7.0.0 (Windows)

* Please uninstall any older SpeechPulse versions and delete the installation folder (e.g., C:\Program Files\SpeechPulse) before installing this new version. During the uninstall procedure, you have the option to keep the downloaded models and CUDA libraries—there is no need to delete them.

• Now supports real-time transcription in live mode.
• Real-time processing significantly improves the accuracy of both Auto and Manual punctuation modes.
• A new option to edit/delete the saved diarization speakers.

9/18/2024 - Version 6.3.2 (macOS)

• Fixed an issue that caused SpeechPulse (macOS) to crash when pressing the caps lock key.

9/17/2024 - Version 6.3.1 (macOS)

• Reduce text hallucinations/repetitions in the Live mode.
• Now automatically re-transcribes any incorrectly transcribed portions of the text in the File mode (e.g., text repetitions/hallucinations).
• A new transcription editor similar to the diarization editor.
• Both transcription and diarization editors now support re-transcribing selected portions of the text. You can select sentences, paragraphs, or any block of text and re-transcribe using the context menu.
• SpeechPulse uses different internal settings for each re-transcription, making it possible to correct any text repetitions/hallucinations.
• You can also restore a previous transcription using the context menu.
• Misc. improvements.

9/17/2024 - Version 6.3.1 (Windows)

• Several performance improvements for the transcription editor to support lengthy audio files.

9/14/2024 - Version 6.3.0 (Windows)

• A new transcription editor similar to the diarization editor.
• Both transcription and diarization editors now support re-transcribing selected portions of the text. You can select sentences, paragraphs, or any block of text and re-transcribe using the context menu.
• Misc. improvements.

9/11/2024 - Version 6.2.0 (Windows)

• Reduce text hallucinations/repetitions in the Live mode.
• Now automatically re-transcribes any incorrectly transcribed portions of the text in the File mode (e.g., text repetitions/hallucinations).
• Also supports manually re-transcribing audio segments in the diarization window. You can use this feature to re-transcribe any incorrectly transcribed segments.
• SpeechPulse uses different internal settings for each re-transcription, making it possible to correct any text repetitions/hallucinations.
• You can also restore a previous transcription for a segment using the context menu.

5/26/2024 - Version 4.2.0 (Windows)

• SpeechPulse (Windows) now supports automatic speaker diarization.
• Supports both CPU and GPU execution for diarization.
• File mode now supports generating all output formats in a single pass.
• A new option to limit the number of words per subtitle line.
• A new indicator to inform users when the push-to-talk mode is active.
• UI improvements.
• Removed the support for onnx models due to their limitations.

5/10/2024 - Version 4.1.0 (Windows)

• SpeechPulse now supports recording your dictations as WAV files inside the "Documents/SpeechPulse" folder.
• You can configure how many days of previous recordings to keep or disable history recordings via "Settings->Options->General settings".
• The new history window allows you to process your previous dictation recordings using different speech models, language models, AI Templates, or any other different settings.
• Now supports importing other compatible offline language models. For example, you can run LLAMA 3 using SpeechPulse. Just place the GGUF language model file inside the SpeechPulse models directory. SpeechPulse will automatically detect the model on startup.
• Fixed an issue that caused SpeechPulse to skip AI formatting when dictated to MS Word, WordPad, and the built-in editor.
• UI changes for better usability on high DPI screens.