InqScribe and Whisper (AI)

InqScribe uses Whisper, by OpenAI, to perform automatic transcription of video and audio files. Whisper is a speech-to-text application that runs locally on your computer.

AI is optional

You have complete control over when and how you use AI. InqScribe only uses AI when you click the "Transcribe" button to generate a transcript. InqScribe does not use AI in any other way.

InqScribe's AI is private

If you do choose to use AI, we designed InqScribe to ensure maximum privacy. Your data never leaves your computer, nothing is transmitted to the cloud, and nothing is shared with OpenAI or Inquirium.

How Whisper works

Whisper uses model files to transcribe audio. Different models will lead to different results. Smaller models are faster but are also less accurate. Larger models will take longer to transcribe and require more computing power, but may be more accurate (not always!).

InqScribe allows you to select from a variety of AI models so you can find the one that performs best for your specific media. InqScribe 3's first-time setup wizard helps you download our recommended models. You can also download additional models from Hugging Face.

We encourage you to experiment with additional models to see if they provide better results.

Performance

The speed of Whisper's automatic transcription depends on your computer's hardware. A more powerful Graphics Processing Unit (GPU) will be able to transcribe audio faster.

If you experience performance issues, try using a smaller Whisper model.

Using your transcript

After Whisper generates your transcript, you can preview it, adjust its format, and when it looks good, add it to an InqScribe document.

InqScribe transcripts are essentially text, so you can easily use them with other tools. For example:

You can copy/paste your transcript into a word processor.
You can export your transcript to SubRip, or another captioning format, to create a video with subtitles.
You can use another another AI tool to generate a summary of your transcript, leveraging your InqScribe timecodes.

Transcription Tips

Here are some tips to help get the best transcription results from InqScribe.

Audio sound quality will affect your transcription. A good microphone and less background noise will help improve transcription accuracy.
Try different models, both larger models and smaller models. The accuracy depends partly on your audio source.

Troubleshooting

I've installed .NET 8 but InqScribe does not recognize it.

If you have installed .NET 8 on your Windows computer but InqScribe does not seem to recognize that .NET 8 is installed, you probably need to install .NET for both ARM64 and x64.

Please visit the InqScribe 3 and .NET 8 page and download and install both of these .NET installers:

Intel CPU (x64)
ARM CPU (arm64)

This is a known issue in 3.0. We will be fixing this in a future release.

I'm getting a Transcription Failed message.

There is a known bug in version 3.0 that results in a Transcription Failed message when using a Windows computer with an ARM processor. e.g.

Transcription Failed
Job Name    declaration_of_independ.mp3 icon
Model   ggml-tiny.bin
Media   declaration_of_independ.mp3 icon
Media Duration  0s
Transcribed at  Jan 9, 2026, 5:26:38 PM
View Raw Output
Close Transcription
Log
Job: transcribe file:///C:/Users/Ben/Downloads/declaration_of_independ.mp3 using model "ggml-tiny.bin"


"C:\Users\Ben\AppData\Local\Programs\InqScribe\InqScribe Resources\wav-trans.exe" "C:\Users\Ben\Downloads\declaration_of_independ.mp3" "C:\Users\Ben\AppData\Local\Temp\rbtFE18.tmp"

You must install .NET to run this application.

App: C:\Users\Ben\AppData\Local\Programs\InqScribe\InqScribe Resources\wav-trans.exe
**Architecture: x64**
App host version: 8.0.19
**.NET location: Not found**

Learn more:
https://aka.ms/dotnet/app-launch-failed

Note the two messages:

Architecture: x64
.NET location: Not found

The workaround is to install .NET for both x64 and the ARM64. Please visit the InqScribe 3 and .NET 8 page and download and install both of these .NET installers:

Intel CPU (x64)
ARM CPU (arm64)

This is a known issue in our current version. We will be fixing this in a future release.

I’m seeing a lot of repetitions in my transcript. Why is that?

This is a known issue with Whisper. While improvements are underway, here are a few things you can try:

Experiment with different models, both smaller and larger. Larger models are not always better; results may vary.
Repetitions are often caused by long pauses in the audio. Try splitting your media into shorter segments or removing long pauses to improve results.

Whisper FAQ

Can InqScribe automatically transcribe my audio and video files?

Yes! InqScribe now includes support for the Whisper transcription model, an advanced AI-based automatic speech recognition (ASR) system.

Are there any per-minute charges for using automatic transcription?

No. Once you have a valid InqScribe license, you can transcribe unlimited minutes at no extra cost.

Is any of my data sent to the cloud for transcription?

No. All transcription is done locally on your computer. Your media stays private, secure, and is never used to train the AI.

What kind of AI does InqScribe use?

InqScribe uses OpenAI’s Whisper. It is a machine learning model, designed for automatic speech recognition (ASR).

Can I choose between different AI models?

Yes. InqScribe allows you to select from a variety of AI models so you can find the one that performs best for your specific media. InqScribe 3's first-time setup wizard helps you download our recommended models. You can also download additional models from Hugging Face.

What are AI models, anyway?

Whisper, the AI used by InqScribe, uses model files to transcribe audio. Different models will lead to different results. Smaller models are faster but are also less accurate. Larger models will take longer to transcribe and require more computing power, but may be more accurate (not always!). We encourage you to experiment with additional models to see if they provide better results.

Are multiple languages supported?

Yes. Whisper supports a number of different languages, including transcriptions of multiple languages in the same transcript, as well as some language translation. See the Whisper GitHub page's Available Models and Languages section.

Can I edit my transcript directly in InqScribe?

Absolutely. You can review, edit, and refine your transcript using InqScribe’s integrated media window and timecode tools.

Can I add notes or comments to my transcript?

Yes. InqScribe offers freeform transcripts, meaning you can add notes or edits anywhere you like—no rigid formatting required.

Is automatic transcription 100% accurate?

No automatic speech recognition system is perfect. See Transcription Tips, above, for more information.

Does InqScribe support speaker identification?

Speaker identification is not supported at the moment. But you might find that some Whisper models will mark voice changes. This is something we are exploring for the future.