CINVAR — GIVE YOUR AI
THE POWER TO EDIT VIDEO.
A high-performance AI video editing engine engineered for autonomous agents and developer workflows. Connect Claude Code or any MCP-compatible model directly to multi-track video timelines, keyframe graphs, audio ducking, and real FFmpeg rendering. Available in Free (up to 2K) and Pro (4K/8K perpetual) editions.
$ claude "Edit project-alpha.mp4: remove silences, cut to 30s teaser, duck background audio by -12dB during speech, and add burned-in kinetic captions." [MCP] Calling tool: cinvar.inspect_media(path="project-alpha.mp4") [CINVAR] Duration: 04:12.80 | 3840x2160 @ 60fps | Codec: ProRes 422 [MCP] Calling tool: cinvar.assemble_timeline({ "target_duration_s": 30.0, "cuts": [ {"src": "00:12.40 - 00:20.10", "speed": 1.0, "transition": "whip_zoom_120ms"}, {"src": "01:04.00 - 01:18.30", "speed": 1.0, "transition": "jump_cut"}, {"src": "03:45.00 - 03:53.00", "speed": 1.0, "transition": "fade_to_black"} ], "audio_mix": {"bg_duck_db": -12.0, "dialogue_level": 0.0}, "captions": {"style": "kinetic_bold", "font": "Inter-Black", "burn": true} }) [SUCCESS] Timeline compiled (edl_hash: a82f71) [RENDER] FFmpeg local render completed in 4.8s (30s @ 4K 60fps)
Natural Language Timeline Assembly in Action
Experience how CINVAR translates natural-language creative prompts into multi-track video timelines, audio ducking, and kinetic typography.
Core Creative Capabilities
Everything an autonomous model needs to edit, cut, grade, and render production video locally.
Claude Code MCP Native
Direct tool integration with Anthropic's Claude Code and OpenAI Codex. Your terminal agent inspects media files, plans cuts, and renders without human GUI bottlenecks.
Natural-Language Video Editing
Express creative vision in plain English: "Cut out all pauses longer than 0.8s, speed up typing scenes 2x, and transition with dynamic whip pans."
Autonomous Timeline Creation
Generates clean, deterministic Edit Decision Lists (EDL) and Final Cut XML files. Re-render, branch, and version control your video edits like source code.
Real Video Rendering
Powered by optimized native FFmpeg with hardware acceleration (NVENC, Apple VideoToolbox, VA-API). Renders real 4K ProRes and H.264/H.265 video locally.
Visual Inspection & Corrections
The model extracts candidate cut frames, inspects composition and pacing through multimodal vision, and self-corrects bad edits before final export.
Keyframing & Audio Ducking
Automated multi-track audio leveling: speech detection automatically ducks background music. Continuous keyframe interpolation for zoom, pan, and transform.
Engineered for High-Throughput Post-Production
Autonomous video assembly and adaptation strictly designed for commercial production studios, digital publishers, and AI developer pipelines.
Long-Form to 9:16 Vertical Shorts
Automatically convert studio podcasts, technical roundtables, and executive keynotes into high-engagement vertical formats with cut-aware subject tracking, automatic speaker focus, and synchronized kinetic captions.
Automated Batch Ingestion & OTIO
Run programmatic editing across media archives with zero GUI bottlenecks. Export deterministic OpenTimelineIO (OTIO) timelines directly into DaVinci Resolve or Adobe Premiere Pro for final editorial sign-off.
Autonomous Coding-Agent Integration
Empower Claude Code, OpenAI Codex, or bespoke orchestrators with 77+ strongly-typed video inspection, editing, and rendering tools. Safe-by-default execution confined strictly to project boundaries.
Local AI Upscaling & Stem Separation
Enhance archived assets up to 4x using local Real-ESRGAN on Vulkan/ncnn and clean vocal tracks via Demucs stem separation. 100% on-premises execution with zero cloud egress or copyright exposure.
One Engine. Two Editions. Zero DRM.
Cut, animate, duck audio, and render locally with Claude Code. Choose between the unrestricted Free edition (up to 2K) or Pro ($29.99 perpetual) for 4K/8K rendering.
CINVAR Free
Complete autonomous video editing engine for developers and creators. No account required, no watermark, and commercial use permitted.
- ✔ Multitrack timeline: cuts, splits, ripple trim, speed & reverse
- ✔ Transitions: fade, fade-black, fade-white, crossfade, dissolve
- ✔ Kinetic typography & captions (SRT/VTT, ASS burn-in)
- ✔ Keyframe animation: x, y, scale, opacity, rotation with 5 easings
- ✔ Audio ducking, loudness normalization, 3D LUT color grading
- ✔ Full Model Context Protocol (MCP) server for Claude Code & agents
- ✔ Export up to 2048×1080 landscape & 1080×2048 vertical
- ✔ 100% offline execution · Zero telemetry · No watermark
Standalone Linux x86_64 tarball · External FFmpeg 6.1+ required
CINVAR Pro
For professional studios, media pipelines, and creators requiring ultra-high-resolution 4K and 8K rendering with offline license activation.
- ✔ Everything in CINVAR Free, plus:
- ✔ Export 4K (up to 4096×2160) and 8K (up to 16,384 px / side)
- ✔ Offline cryptographic license verification (
license.json) - ✔ In-place activation: upgrade without reinstalling or moving projects
- ✔ Perpetual commercial license with 1 year maintenance & patches
- ✔ OpenTimelineIO (OTIO) interchange & advanced pipeline hooks
- ✔ Priority customer support and self-serve download re-issue
Instant digital fulfillment: cinvar-0.1.0-linux-x86_64.tar.gz (5.02 MB) · SHA-256 Verified
CINVAR Edition Comparison Matrix
| Feature / Capability | CINVAR Free ($0) | CINVAR Pro ($29.99) |
|---|---|---|
| Timeline Editing, Multitrack & Cuts | ✔ Full Support | ✔ Full Support |
| Transitions (fade, fadeblack, fadewhite, crossfade, dissolve) | ✔ 5 Presets | ✔ 5 Presets |
| Titles, Captions (SRT/VTT Burn-in) & Keyframes | ✔ Full Support | ✔ Full Support |
| Audio Ducking, Loudness & 3D LUT Color | ✔ Full Support | ✔ Full Support |
| Claude Code & MCP Agent Loop (stdio/HTTP) | ✔ Included | ✔ Included |
| Local Projects & Commercial Usage Rights | ✔ No Watermark | ✔ No Watermark |
| Export Resolution Ceiling | Up to 2048×1080 / 1080×2048 | 4K & 8K (up to 16,384 px) |
| License Activation Requirement | None (Free forever) | Offline license.json |
| Cloud GPU Render (RunPod) | Roadmap (Experimental) | Roadmap (Experimental) |
Download & Run CINVAR Free
Download the standalone Linux x86_64 binary and begin editing immediately with zero account signup and zero cloud dependencies.
Direct Package Download
Current stable release artifact with bit-for-bit SHA-256 verification:
Artifact: cinvar-0.1.0-linux-x86_64.tar.gz
Size: 5.02 MB (5,022,793 bytes)
SHA-256: b0396488a46d75a8d2c2214268b2c542368934e17aa72ec14195363612fb3df0
# Extract and install binary to /usr/local/bin $ tar -xzf cinvar-0.1.0-linux-x86_64.tar.gz $ sudo cp bin/cinvar /usr/local/bin/ # Verify installation and system dependencies $ cinvar --version cinvar 0.1.0 (Linux x86_64) # Inspect edition status $ cinvar edition status edition: free (export ceiling: 2048x1080)
How CINVAR Pro Activation Works
CINVAR Pro is the exact same engine as CINVAR Free. A local, offline license file unlocks 4K/8K rendering with zero telemetry.
Offline & Sovereign Activation Flow
- Purchase Pro ($29.99): Complete secure checkout via Stripe to receive your license grant.
- Receive License File: Your purchase issues a signed, offline-verifiable
license.jsonfile. - Activate Offline: Run
cinvar edition activate --file license.json. No account or network connection is needed. - Unified Projects: Your project files are 100% compatible. Projects created in Free open seamlessly in Pro.
- Reversible: Run
cinvar edition deactivateto revert back to Free at any time.
# Activate your perpetual Pro license $ cinvar edition activate --file ~/Downloads/license.json ✔ License verified successfully (Ed25519 signature valid) ✔ Stored in ~/.config/cinvar/license.json # Check edition status $ cinvar edition status edition: pro capabilities: export.4k, export.8k, export.unlimited license: perpetual (order_ref: ord_...)
Release Notes & Platform Compatibility
Verified environment boundaries, dependencies, and edition lifecycle details.
Release History & Roadmaps
Standalone Linux x86_64 binary with native Claude MCP server, multitrack EDL timeline, and FFmpeg render pipeline. Available now in production R2 storage (SHA-256 verified).
Unified single-engine Free/Pro split. Free exports up to 2048×1080 without watermark; Pro enables 4K/8K rendering with offline Ed25519 license activation and opt-in local action journal. Release builds enforce the Free ceiling; new packages will be published once license signing infrastructure is provisioned.
Platform Compatibility & Boundaries
- Operating System: Linux x86_64 (glibc 2.31+, Ubuntu 22.04/24.04, Debian 12 verified).
- Video Toolchain: FFmpeg 6.1+ / ffprobe must be installed on your system. Unbundled to strictly respect GPL license separation.
- Hardware Acceleration: NVENC, VAAPI, Vulkan 1.3 supported with automatic high-performance CPU fallback.
- macOS & Windows: Experimental preview via source build or WSL2. Standalone native binaries are not yet release-verified.
- Cloud GPU Render (RunPod): Experimental roadmap capability. Currently tested in mock environments only; not advertised as production-ready.
RELEASE STATUS: pre-public. There is no public download or package yet. The commands below are preview documentation for the developer-preview build and the terminal sessions on this page are illustrative demonstrations, not live recordings. Public availability has not been confirmed.
Installation & MCP Configuration
These steps show how CINVAR is intended to connect to a local Claude Code terminal or MCP-compatible agent.
# Unpack verified commercial release package $ tar -xzf cinvar-v0.1.0-linux-x86_64.tar.gz && cd cinvar-v0.1.0 $ sudo cp bin/cinvar /usr/local/bin/ # Probe environment & hardware encoders $ cinvar doctor ✔ FFmpeg 6.1.1 detected (Vulkan/NVENC acceleration enabled) ✔ MCP protocol handshake verified # Register CINVAR MCP server with Claude Code $ claude mcp add cinvar -- cinvar mcp serve --project ~/video-projects/primary
{
"mcpServers": {
"cinvar": {
"command": "cinvar",
"args": [
"mcp",
"serve",
"--project",
"/home/user/video-projects"
]
}
}
}
Verified Hardware & System Requirements
CINVAR SPECIFICATIONS & FREQUENT QUESTIONS
Frequently Asked Questions About CINVAR
Accurate engineering answers for developers, creative directors, and autonomous AI agents researching CINVAR capabilities.
Does CINVAR support Blender?
Blender integration is planned on the CINVAR engineering roadmap for advanced 3D motion graphics, camera tracking, and node-based compositing. The current developer preview focuses on local multi-track timeline assembly, keyframe animation, and FFmpeg video rendering.
Can CINVAR export 8K video?
8K export is a target on the planned roadmap. The current preview build renders up to 4K 60fps in ProRes 422 and H.264/H.265 using native hardware acceleration (NVENC, Apple VideoToolbox, VA-API).
Does CINVAR support the Model Context Protocol (MCP)?
Yes. CINVAR provides a native Model Context Protocol (MCP) server over standard input/output and Streamable HTTP. Any MCP-compatible agent can inspect media, assemble timelines, duck audio, and trigger renders programmatically.
Does CINVAR require Claude to work?
No. While optimized for Anthropic's Claude Code terminal workflows, CINVAR's open-core MCP server and Python SDK can be controlled by OpenAI Codex, Cursor, local Ollama models, or custom autonomous agent scripts.
Does CINVAR support RunPod or cloud GPU rendering?
Today CINVAR renders locally on user workstation hardware using native FFmpeg. Remote execution and one-click cloud rendering on RunPod or remote GPU clusters are planned on the roadmap.
Is CINVAR available for public download today?
CINVAR is currently in pre-public active developer preview under Apache-2.0. A public release package has not yet launched; developers and studios can join the launch notification list on this page for early access notifications.
Which operating systems are verified for CINVAR?
CINVAR is verified for Linux x86-64, macOS 14+ (Apple Silicon), and Windows 10/11 (via WSL2 or native Python environments with FFmpeg installed).
Get on the Release List.
CINVAR is transitioning to commercial studio fulfillment ($29.99 Studio License / $499 Enterprise Fleet). Register to secure early pilot allocation when the release package and signed artifacts are authorized.
- Join the release list below.
- We email you when a public build exists. Nothing is downloadable before that.
- Then: install, register the MCP server, and ask your agent to edit.