jasna: A Python Video Restoration Tool

TL;DR

TL;DR: jasna is a video restoration tool written in Python and released under AGPL-3.0. It has 2,442 stars and 189 forks after roughly nine months, which is fast growth for a single-purpose media tool.

What it is

The repository description is five words long: video restoration tool. That is the whole thing, and it is worth being upfront that there is not more to it. Everything below is grounded in what the GitHub API reports about the repository, not in claims about internal architecture.

Here is the metadata that is actually verifiable:

Field	Value
Repository	Kruk2/jasna
Stars	2,442
Forks	189
Open issues	62
Language	Python
License	AGPL-3.0
Created	2026-01-18
Last push	2026-10-18
Homepage	None listed
Topics	None listed

Two numbers stand out. The first is the star to fork ratio. At roughly thirteen stars per fork, this is a repository people are watching and recommending rather than one people are immediately forking and reshaping, which fits a tool with a clear single job. The second is the open issue count. Sixty-two open issues on a 2,400-star repository is an unusually high load, and it is the clearest signal here that the project is generating real usage faster than its maintainer capacity absorbs it.

Where this category sits

Video restoration covers the work of taking degraded footage and putting it back into a usable state. In practice that umbrella spans a few distinct problems:

• Upscaling footage that was captured at a lower resolution than you need
• Recovering detail that compression or transmission artefacts smoothed away
• Handling per-frame flicker and temporal inconsistency that per-frame processing introduces
• Working with formats that decoders rarely handle well

Tools in this space usually have to deal with that third item, because restoring frames independently is easy and produces a subtle result that people notice immediately. Consistency across neighbouring frames is what separates a restoration that looks acceptable from one that looks like it is boiling.

Python being the language of record for this category is not a coincidence. The tooling ecosystem around neural restoration is Python-first, so a Python tool sits where the models and the converters already are.

The licensing question

AGPL-3.0 is the part of this repository worth slowing down on. It is the strongest copyleft of the common open source licences, and unlike GPL-3.0 it carries network use into scope. If you run a modified version of jasna as a service and let other people reach it over a network, you are required to offer those people the corresponding source.

That is fine for personal use, for research, and for internal tools. It matters if you were planning to build something commercial on top of it without releasing your changes. Check your own situation before committing to it.

When to use it

It fits the case where you have footage worth recovering and you want a tool rather than a platform. The star count suggests a real community already behind it, which matters for a project maintained by one person.

When not to use it

Two cautions. First, sixty-two open issues is a lot of unanswered surface for a repository this size, so if your work depends on a specific fix or a specific feature, factor in that you may be waiting. Second, with no listed homepage and no listed topics, there is little outside the repository to tell you how the tool is meant to be used. Read the repository before you plan around it.

Source and Accuracy Notes

• Project page: https://github.com/Kruk2/jasna
• Source repository: Kruk2/jasna (verified via GitHub API html_url)
• License: AGPL-3.0 (verified via GitHub API license.spdx_id)
• Stars at time of writing: 2,442
• Repository description as published: “Video restoration tool”
• The repository lists no homepage and no topics in its GitHub metadata as of this writing.