Agent workflow
01Edit through conversation
Connect compatible ACP agents, inspect timeline state, and move from intent to concrete edits without leaving the editor.
Source-only alpha
Video editing with an agent in the timeline.
Anica is an agentic-first video editor where AI is a collaborator, not a side panel. Talk to an agent, inspect your timeline, clean speech, translate subtitles, and drive media workflows from one native Rust application.
preview
Rough cut with agent notes
PLAY
T1
T2
T3
T4
T5
agent prompt
Help me cut silent parts, remove duplicate speech, then suggest B-roll for the open gaps.
planned actions
- Analyze audio energy
- Mark repeated speech
- Draft semantic layer ideas
export path ready
Core
Rust + GPUI + WGPU
Media
FFmpeg
Intelligence
ACP + LLM workflows
Why it exists
AI should understand the edit, not just generate assets.
Traditional editors bolt AI onto separate dialogs. Anica makes the agent part of the editing model, so natural language can map to timeline-aware actions.
Subtitle pipeline
02Local speech and translation paths
Use local ONNX Whisper model packs for subtitle generation, then translate and re-import subtitle tracks with timing preserved.
Native media core
03Built for GPU-assisted editing
Rust, GPUI, WGPU, GStreamer, and FFmpeg provide the foundation for native preview, export, analysis, and effect rendering.
Editing loop
Keep creative control while the agent handles the mechanical passes.
The goal is not one-click editing. The goal is a faster conversation between your intent, the timeline, and the media engine.
1
Import clips and build a timeline.
2
Ask an agent to inspect, clean, or restructure the edit.
3
Generate subtitles, translations, B-roll ideas, or export actions.
4
Review the timeline and keep creative control over the result.
Motion Graphics & VFX
Generate motion graphics with natural language.
Anica includes MotionLoom, a GPU-powered motion graphics and VFX runtime built on WGPU. It supports timeline-aware compositing, animated text, overlays, transitions, and procedural effects — all drivable by AI agents from natural language prompts.
prompt → motionloom dsl
Create an 8-second cyber-style motion graphic using the MotionLoom DSL: black background, white text, typewriter-style reveal for the word "ANICA", with subtle floating movement and slow drifting motion in the air.
The prompt generates MotionLoom DSL code, which you can copy and paste on any compatible computer to reproduce the same motion graphics and VFX.Open source
Built in public for agent-native creative tools.
Anica is currently alpha software with source-only distribution. Contributions, experiments, and workflow feedback are welcome.