Download SpeechPulse

Try SpeechPulse with our 30-day free trial.

Latest Windows version:

∗ This version of SpeechPulse requires newer CUDA libraries. Please download the newer CUDA libraries from here.

∗ When running the latest SpeechPulse installer on Windows, you may get a SmartScreen warning message. This warning usually occurs shortly after a new release of SpeechPulse until it gets enough reputation. You can continue to install SpeechPulse by clicking the "More info" link in the SmartScreen warning dialog.

∗ SpeechPulse is 100% safe and does not contain any malware. If your computer displays any warnings, please report it as a "false positive" (incorrect detection) to your virus guard manufacturer.

Older versions:

System requirements

For Windows:
  • Windows 10 64-bit (version 1809 or later) or Windows 11
For macOS:
  • An Apple silicon Mac
  • macOS 13+

How to install SpeechPulse on macOS (Apple silicon)

Speech Profiles

If you are in a noisy environment and SpeechPulse generates random text, you can use the "Speech Profiles" feature to add a new speech profile that only detects your own voice.

If multiple users share the same SpeechPulse installation, each user can create a separate speech profile sensitive to their voice.

The speech profiles feature can ignore noisy background audio and eliminate text hallucinations. However, it can't separate speech from multiple active speakers.

How to add speech models - Windows

SpeechPulse installer only includes the "English (base) L" speech model. For better accuracy and multi-language support, you can download additional speech models listed below.

Larger speech models have better accuracy than smaller models, but smaller models run faster and have lower processing times.

We highly recommend running larger speech models (Multi (large) L, Multi (medium) L, and English (medium) L) on an NVIDIA GPU for faster live transcription. These models can be too slow on a CPU for voice typing.

For example, the Multi (large) L model takes less than 2 seconds to transcribe the sentence "Speech recognition is much faster than typing." on an NVIDIA RTX 3060 GPU. The same sentence takes 16 seconds on an Intel Core i5-12400 CPU.

Speech models - Windows

These models can run on both the CPU and the GPU. They give the fastest results with CUDA-enabled NVIDIA GPUs.

Accuracy for English language:

Multi (large) L > English (medium) L > English (small) L > English (base) L > English (tiny) L

Accuracy for other languages:

Multi (large) L > Multi (medium) L > Multi (small) L > Multi (base) L > Multi (tiny) L

You need to unzip the downloaded models into the "models" directory inside the SpeechPulse installation directory (e.g.: "C:\Program Files\SpeechPulse\models"). Next time you run SpeechPulse, it will automatically detect the new models.

CPU and GPU models

* To run these models on the GPU, you need an NVIDIA GPU with CUDA support. You also need to download one of the CUDA library zip files below according to your SpeechPulse version. Then unzip the zip file to the SpeechPulse installation directory (e.g.: "C:\Program Files\SpeechPulse").

* If you already have older CUDA libs on the SpeechPulse installation directory, please replace them with the new libraries after installing SpeechPulse version 4.2.0 or later.

CUDA libs

How to add AI text-processing language models - Windows

You can use SpeechPulse to enhance, summarize, and format your dictated text using the AI language models provided below.

We highly recommend running these language models on an NVIDIA GPU for faster results. They can also run on a CPU but can be too slow for live transcription.

However, if you don't want real-time results (e.g. File mode), you can run these on a CPU.

You need to place the downloaded models into the "models" directory inside the SpeechPulse installation directory (e.g.: "C:\Program Files\SpeechPulse\models"). Next time you run SpeechPulse, it will automatically detect the new models.

AI language model location Windows

* Make sure your PC has sufficient RAM/VRAM to run the AI language model alongside the speech model. You can run both models on the CPU, both on the GPU, or one on the CPU and the other on the GPU. For example, if you want to run the "Multi (large) L" speech model and "English (standard)" language model, both on the GPU, you need at least 8GB of VRAM (~3GB for "Multi (large) L" speech model + ~4GB for "English (standard)" language model).

How to add speech models and AI text-processing language models - macOS

SpeechPulse macOS version has a built-in model downloader. You can use it to download both speech models and language models.

How to get better accuracy

1) Try to reduce the background noise

SpeechPulse has decent accuracy even with moderate background noise. However, for the best accuracy and lowest latency, you should try to lower the background noise as much as possible.

2) Try to speak in complete sentences

Short phrases can confuse the AI speech models, causing poor accuracy. So try to speak in complete sentences as much as possible.

3) Use a headset microphone instead of a PC/laptop microphone

PC/laptop microphones can capture faraway sounds/noises causing random text hallucinations in SpeechPulse. So it is better to use a headset microphone for better accuracy and fewer hallucinations.

4) Use a larger speech model

Larger speech models have better accuracy. They also work well under background noise. However, larger models require more RAM and have higher latencies.