EditPostFX

Category

Local AI

On-your-machine video, voice and enhance tools. A serious GPU often matters.

Whisper

Open source

OpenAI's local speech-to-text. The code is MIT. You still need to read the model terms, and a GPU makes life much easier.

Windows macOS Linux
MIT (code). Check model terms. Official site

WhisperX

Open source

Whisper with better alignment and speaker labels. A strong local path toward Descript-style captions. Still not a full text editor NLE.

Windows macOS Linux
BSD-2-Clause Official site

Faster-Whisper

Open source

A faster reimplementation of Whisper. Same idea, usually quicker on the same machine.

Windows macOS Linux

Piper

Open source

Local text-to-speech you can run on your own machine. Do not clone someone else's voice without clear consent.

Windows macOS Linux
MIT (code). Voice models have their own terms. Official site

ComfyUI

Open source

A local node-based interface for image and video models. A serious GPU is effectively required, and quality still lags the best paid video generators.

Windows macOS Linux
GPL-3.0 (app). Model licences vary and can be restrictive. Official site

Real-ESRGAN

Open source

A widely used open upscaler for stills and frames. Useful, not a complete Topaz Video AI substitute.

Windows macOS Linux
BSD-3-Clause (code). Check model terms. Official site

Video2X

Open source

A video upscaling and frame-interpolation frontend that wraps several open models. GPU recommended.

Windows Linux
AGPL-3.0 Official site

RIFE

Open source

Open frame interpolation for smoother slow motion. Results vary a lot by footage.

Windows macOS Linux
MIT (code). Check model terms. Official site

Waifu2x

Use with care

An older open upscaler that still helps on some stills, especially illustrations. Not a modern general video enhancer.

Windows macOS Linux

ProPainter

Restricted licence

Research code for video inpainting and object removal. Quality is uneven, and you must check the licence.

Windows macOS Linux
S-Lab Licence 1.0 (non-commercial unless you get permission) Official site

SAM 2

Open source

Meta's open segmentation model, useful for masks and object isolation in video pipelines. You still have to build the workflow around it.

Windows macOS Linux
Apache-2.0 (code and stated model terms: check the repo) Official site

DeepFilterNet

Open source

Local speech denoising that can help rough dialogue. It will not match iZotope RX on the hardest cleanups.

Windows macOS Linux
MIT (code). Check model terms. Official site

RNNoise

Open source

A well-known open noise-suppression library, often used inside other tools. A helper, not a full restoration suite.

Windows macOS Linux
BSD-3-Clause Official site

whisper.cpp

Open source

A fast C/C++ port of Whisper that runs well on a CPU, and even better on Apple Silicon. Same models, much easier to run locally.

Windows macOS Linux
MIT (code). Check Whisper model terms. Official site

Demucs

Slow maintenance

Splits a music track into stems such as vocals, drums and bass. The code is MIT. The official weights are provided for research use only, and the project is only lightly maintained.

Windows macOS Linux
MIT (code). Official model weights are research-only, not MIT. Official site