VideoProc Converter AI

Overview

Most video work that is not editing is one of five jobs: change the format, make the file smaller, trim and crop it, record what is on the screen, or pull a video off a disc. VideoProc puts all five behind five buttons on one screen and does all of them competently, which is why it stays installed on machines that also have a full edit suite.

The AI in the name refers to a set of trained passes bolted onto the converter: upscaling old or low resolution footage, interpolating frames for smooth slow motion, stabilising handheld recordings, and a general restoration pass for compressed material. This build is posted with the licensing handled, so conversions run at full length rather than stopping at the five minute trial limit, and output carries no watermark.

Conversion and compression

The format list is long and organised by device and destination as well as by codec, so you can pick what the file is for rather than working out a bitrate. Hardware acceleration runs on NVIDIA, Intel and AMD paths and is the reason a full length conversion takes minutes here where a software encoder takes an hour.

Compression is handled as its own mode, with a target size or a quality level and a preview of the result at the chosen setting before you commit. Batch conversion queues a folder and processes it unattended, and the queue survives being paused and resumed.

The machine learning passes

Super resolution upscales, and unlike a simple resize it reconstructs detail. It is at its best on clean but low resolution source, an old digital camera video or a downloaded clip, and less good on heavily compressed material where it sharpens the compression blocks along with everything else. Two models are provided, one tuned for real footage and one for animation, and picking the right one matters more than the strength setting.

Frame interpolation generates intermediate frames, which serves two purposes: raising an old 24 frame per second clip to something smoother, and producing slow motion from footage that was not shot at a high frame rate. Fast movement across the frame is where it shows its seams.

Stabilisation analyses the motion across the clip and reframes each frame to cancel the shake, cropping slightly in the process. On moderately shaky handheld footage the improvement is large. The restoration pass cleans up noise and compression artefacts and is best used gently.

Editing, recording and ripping

The editing tools are deliberately basic and that is fine: trim, cut, merge, crop, rotate, adjust levels, add a subtitle track, apply a simple effect. It is faster than opening a real editor when the job is cutting thirty seconds off the front of a clip.

The recorder captures the screen, a window, a camera, or a combination with picture in picture, with system audio and microphone on separate tracks. There is a drawing overlay for annotating during a capture, which makes it usable for putting together a quick demonstration.

Disc ripping handles video discs including the protected ones, with output either as a full copy or as a converted file. Analogue titles with deliberately broken structures are handled by a repair pass that many rippers still fail on.

Performance notes

Ordinary conversion is fast on any machine with a recent graphics card because the encoding happens on the card. The machine learning passes are a different matter, since they are inference work and run at a few frames per second on a mid range card. Upscaling a two hour film is an overnight job, not a lunch break.

The program shows a time estimate before starting, and processing can be paused and resumed. Running an upscale and a frame interpolation as two separate passes takes longer but gives noticeably better results than asking for both at once.

What you get

• Format conversion with device and destination organised presets
• Hardware acceleration on NVIDIA, Intel and AMD
• Compression mode with target size and preview
• Super resolution upscaling with real footage and animation models
• Frame interpolation for smooth motion and slow motion
• Stabilisation for handheld recordings
• Noise and compression artefact restoration
• Basic trim, cut, merge, crop and subtitle tools
• Screen, window and camera recording with separate audio tracks
• Drawing overlay during capture
• Disc ripping including protected and damaged structures
• Batch queue that pauses and resumes

Inside the archive

• VideoProc Converter AI, offline installer
• Machine learning model set, preinstalled
• Recorder and ripper components
• Interface language packs
• Readme with a note on model processing speed

System requirements

Installing it

1. Unpack the archive to a local folder.
2. Disable real time protection before running setup.
3. Install and let it finish without launching.
4. Apply the licensing step from the readme if setup did not.
5. Convert a short clip and check the output for a watermark before starting anything long.

Mirrors

Release history

Questions about this release

Comments

Comments are read before they appear. If a build stops working, say so here and it gets rebuilt rather than quietly left up.