An AI agent that operates After Effects

What the agent does inside After Effects

The After Effects panel is a bidirectional bridge into your comps. The agent controls composition state, reads back the properties on any layer, sets layer properties, places markers, and queues renders, then reports the queue status to you. This is the motion-graphics and finishing end of the same agent that assembles your edit, so a comp can react to selects, markers, and timings the agent already understands from your footage.

• Composition control

  Operate comps and their state from the panel.

• Layer properties

  Read and set properties on individual layers.

• Markers

  Place and label markers across compositions.

• Render queue

  Queue comps to render and report status back.

• Vision analysis

  Inspect frames for text overlap, framing, and overlays.

• Colour matching

  Match graphics to a graded plate or reference.

How the panel works

EditAssist runs as a single agent on your machine. The CEP panel connects that agent to After Effects both ways, reading composition and layer state and sending control operations back. You install the panel once and dock it beside your comps, and the conversation you have with the agent operates the project in front of you.

The same agent is reachable from the EditAssist desktop app, the terminal, or a Telegram message, sharing the same project context and memory. After Effects and Premiere Pro share the Adobe CEP bridge, so the agent moves between editorial and graphics without losing the thread.

Commands you can give it

You work by describing the outcome. A few After Effects-specific examples:

• “Read the layer properties on the title comp and tell me which are keyframed.”

• “Add a marker on the main comp at every beat drop in the music track.”

• “Set the opacity on the lower-third layer to fade in over the first 12 frames.”

• “Queue this comp for render at full resolution and report when it's done.”

• “Build a contact sheet of keyframes from the montage so I can pick stills.”

• “Match the colour temperature of the graphics to the graded plate.”

• “Find every shot where on-screen text overlaps the action and flag it.”

• “Generate a before-and-after split screen of the VFX shots for the supervisor.”

Hundreds more across every workflow live in the prompt library.

Graphics, finishing and render

After Effects is where titles, motion graphics, and VFX finishing happen, and EditAssist's vision and timing understanding feed straight into it. With every frame indexed and every word transcribed already, the agent can place markers at beats or topic changes, find frames where graphics overlap the action, build before-and-after comparisons for supervisor review, and match the colour of graphics to a graded plate.

When a comp is ready, the agent queues it to the render queue and reports back. It fits into a pipeline that also reaches DaVinci Resolve for grading and conform, so editorial, graphics, and finishing share one agent and one indexed understanding of the footage.

Local-first by design

Transcription, vision analysis, embeddings, and search indexing all run on your own GPU. No frames are uploaded and no media is sent to a server. The only thing that leaves your machine is the text of your conversation with the agent. Your client's footage and your project files stay on your hardware. EditAssist runs on macOS (Apple Silicon, Metal) and Windows (x64, CUDA).