Voice-to-text transcription

OpenAI’s Whisper voice-to-text model is open source and you can run it on your own PC with tools like Whisper Desktop. Whisper Desktop will run the Whisper model on your PC’s GPU for fast transcription.

It’s a capable solution for converting audio to text. You can speak directly into your microphone or provide an audio file. While Whisper isn’t perfect—no AI tool is—it does outmatch the professional transcription software you would’ve had to pay for just a few years ago.

Image upscaling

These days, so many companies have caught up on offering cloud-based image editing and upscaling tools. Adobe Photoshop even has this feature, but Photoshop does it on Adobe’s cloud servers.

If you want to increase the resolution of images using your own PC, Upscayl is a user-friendly tool for upscaling images from lower resolutions to higher ones via local AI.

Cloud-based AI image editing tools are convenient, but if you have a powerful enough rig, this is the type of thing you can do right on your PC without uploading your images to a cloud server.

Real-time webcam and microphone effects

Microsoft is really pushing Windows Studio Effects as part of its Copilot+ PC suite of AI features, and many of the latest laptops I’m reviewing have “AI webcam effects” packages preinstalled. If you have a Copilot+ PC laptop, try using Windows Studio Effects. If you have a recent laptop in general, dig in the Start menu for webcam filter tools.

But if you have a powerful gaming PC (whether a desktop or laptop) with an Nvidia RTX GPU, you can use the free Nvidia Broadcast app to unlock AI webcam and microphone effects like background removal, fake eye contact, and even high-end features like “studio-quality lighting” on top-end GPUs. It all happens in real time, so you can use it while live-streaming a game or in a video meeting.

Video upscaling and editing

You can AI upscale and edit videos using your PC’s own hardware, too. Topaz Labs offers popular paid professional apps for AI video and image editing work, with all the processing happening on your PC’s local hardware. It’s a pricey solution designed for professional workflows, but it shows what’s possible with local AI.

For a free and open-source option, take a look at Video2X. That one’s a surprisingly user-friendly tool for AI-upscaling video files.

These tools are good examples of the “last mile” challenge. While there are lots of powerful local AI models out there, the most polished user interfaces that are easy to work with tend to be paid tools. Hobbyists and researchers can make powerful software, but they often don’t spend much time on polishing it into a shiny end-user product.

Voice cloning

Did you know you can clone your voice using your PC’s hardware? Tools for this aren’t particularly polished yet—like lots of the local AI landscape—and you’ll often get a web UI and have to download some large files. You can do this with GPT-SoVITS or RVC, but expect some jankiness.

However, it’s a great example of what’s possible: you can already clone a voice using consumer hardware and some open-source software. The only missing piece of the puzzle is an easy user interface.

Music generation

If you’ve seen AI-generated songs on social media, they were probably created using Suno, a cloud-based music generation tool.

Local AI solutions for generating music exist, but most of them are early in development and still unpolished. YuE is an open-source tool that looks like it could one day compete with Suno. You can download YuE and run it on your own hardware, but you’ll probably want to stick with Suno until tools like YuE are more user-friendly.

As is often the case with local AI solutions, YuE is making it easier to access the kinds of features that were only available via companies running on cloud servers in the past. According to YuE, generating 30 seconds of audio takes about 360 seconds (6 minutes) on a PC with an RTX 4090 GPU. That’s not bad! Give it a few more years and you might be able to generate full songs on your gaming PC.

Remove vocals from music

If you like to perform karaoke to backing tracks, or if you just prefer to listen to instrumental music, you may wish you had a tool that could remove the vocals from any song. People have been able to do that for a long time, but it’s been a painstaking process that takes a lot of time—until now, thanks to Ultimate Voice Remover.

This free application is simple, user-friendly, and gets the job done in mere minutes rather than hours or even days. Just provide an MP3, FLAC, or WAV file and it’ll spit out a version with vocals stripped.

Local AI is powerful but unpolished

If you’ve been disappointed by the amount of AI hype over the past few years, I understand. Despite all the high-flying talk about local AI, Microsoft Windows and consumer software packages have done very little integration of useful AI tools.

The most interesting things are happening in the open-source software community, where surprisingly powerful local AI models come with unsurprisingly janky and amateurish user interfaces. Fortunately, there’s a good chance more user-friendly solutions will pop up in the next few years that take better advantage of powerful PC hardware.

For now, you can already do a lot with local AI if you’re willing to get your hands dirty, suffer through rough learning curves, and equip yourself with some relatively powerful hardware (e.g., RTX GPU). Unfortunately, NPUs won’t help you run local AI tools just yet.

Want more PC goodness? Sign up for Chris’s newsletter, The Windows ReadMe. It’s always written by a human, even when it’s about AI.