7 Open-Source Repos to Replace After Effects? The Comments First Ask Whether It Can Keyframe

On X, a user named eigen posted that he no longer uses After Effects. He said he was down to just two things: an AI model called Opus 5.5 and an open-source toolkit.

He listed 7 GitHub repositories covering the entire chain from idea to finished video.

Specifically, these 7:

01 Remotion, Video Engine
Write videos with React, rendering visuals as code. Suitable for batch output, such as generating one short video each for thousands of products. The README clearly states that it has its own license, and some company use cases require purchasing a separate commercial license.
github.com/remotion-dev/remotion

02 Motion Canvas, Animation Workbench
Aimed at informational vector animation, using TypeScript to write animation logic, with real-time preview. It's handy for explainer videos and process demonstrations.
github.com/motion-canvas/motion-canvas

03 Blender, 3D Studio
Modeling, rigging, animation, simulation, rendering, compositing, and video editing are all included. A longtime staple among open-source 3D tools.
github.com/blender/blender

04 Natron, Compositing Lab
A node-based compositing tool for keying, tracking, and multi-layer compositing. It targets software like Nuke.
github.com/NatronGitHub/Natron

05 ComfyUI, Generation Pipeline
Operate generative models through a node-based interface, producing both images and video assets.
github.com/Comfy-Org/ComfyUI

06 Kokoro, Voiceover Studio
An open-source text-to-speech model responsible for narration.
github.com/hexgrad/kokoro

07 FFmpeg, Post-Production Tool
Transcoding, muxing, and compression. The final step is unavoidable.
github.com/FFmpeg/FFmpeg

The workflow he gives is straightforward: idea, scene, effects, mixing, export.

The comments split into several camps. Someone asked whether it can keyframe layer by layer, and eigen replied yes. Someone asked how to set up a stack themselves; his answer was: send this post to your agent, have it download the repositories, and when generating video, have the agent use these tools. Someone else directly asked why there's no hyperframes; he didn't reply to that one.

A complete toolchain doesn't mean you can produce a finished video. Remotion requires knowing React, Motion Canvas requires knowing TypeScript, and Blender's learning curve can scare off plenty of people. eigen says he no longer uses AE, most likely because he already has a development background.

The really interesting part of this combination isn't saving money. It's that it turns video production from opening software and dragging a timeline into writing code, issuing commands, and waiting for renders. The AI agent handles orchestration, while the CLI and API handle execution.

For people without a development background, the barrier isn't the price of software; it's whether they can translate the visuals they want into executable instructions. The question about keyframes in the comments is really asking exactly that.