Skip to content
CoClip
Download
Windows & macOS

Edit short videos with your AI agent.

CoClip is a desktop editor for TikTok, Reels and Shorts where your AI agent edits the timeline with you — live, labelled and one click from undo.

Download for WindowsWatch demo

Windows 11 & macOSNVIDIA-accelerated export on WindowsOn macOS? Get that version

Animated demo: you ask your AI agent to make the hook pop and add captions; the timeline and the 9:16 preview update live.

Short videos are quick to watch. Not to edit. CoClip turns the edit into a conversation — and keeps you in control.

How it works

You talk. Your agent edits. You decide.

Start from an AI draft, ask in plain language, and watch every edit land on the timeline.

Works with
and other MCP-compatible agents
3 edits by Claude

CoClip editor showing the current step

  1. An AI pipeline — or your agent — builds the first cut: clips, voice-over, a word-level transcript, captions and titles.

  2. Say “move the voice 0.5 s later.” Claude, Codex or Gemini edits through CoClip’s built-in MCP server — or open the Claude panel right inside the editor.

  3. Every change appears on the timeline as it happens and is labelled in the AI log — by agent, when several build one video at once.

  4. Any edit your agent makes can be undone in one click, just like your own.

  5. Export from 480p to 4K, accelerated by NVIDIA GPUs on Windows. What you see is what you export.

  • Every AI edit is labelled

    Agent changes are marked on the timeline and listed in the AI log, so you always know who changed what.

  • One-click undo

    Any edit an agent makes can be undone in one click, just like your own.

  • Unsafe edits are rejected

    Edits that would break your project are refused, and the editor and your agents never overwrite each other — even with several agents at work.

  • Editor open or closed

    Your agent can work on a project whether CoClip is open or not. When it is, you watch the edits land live.

Captions

Captions that get Vietnamese right.

Word-highlight captions from a transcript made on your PC, with rounded boxes, stroke, shadow and emphasis colours. Vietnamese diacritics render perfectly.

Phone preview: a Vietnamese caption highlighted word by word.

Performance

Export faster than you can watch.

CoClip’s own export engine composites every frame on the GPU and encodes with NVIDIA NVENC when you have it. Other GPUs, or none at all, run the same engine with a software encoder, so the output stays the same.

About 6.5× faster than real time on NVIDIA GPUs

One short · 1080 × 1920 · 30 fps · multi-layer

NVENC · RTX 5060 Ti
Export timeabout 11 seconds
Video length71 seconds
~6.5×faster than real time on this test

Measured on the developer’s machine with an RTX 5060 Ti: a 71 s short with video, images, stickers, cards and text exported in ~11 s (median of three cold runs). Your results will vary with hardware and project.

Export options
Resolution480p720p1080p2K4K
Frame rate242529.97305059.9460 fps
CodecH.264H.265
FormatMP4MOV
Bit rateLower, Recommended or Higher, plus your choice of output folder
AudioLoudness normalized to −14 LUFS, ducking under voice
EncoderNVIDIA NVENC on Windows when available · otherwise GPU compositing with a software encoder · software route with no GPU · [CONFIRM MAC ENCODER]

Features

A complete editor, before you ask an agent.

Everything here ships today. Hover a tile to see it move.

  • A multi-track timeline that feels familiar

    Video, image, overlay, text, caption, audio, SFX, element, filter and camera tracks. Drag from the media library, snap, multi-select, split, trim, ripple delete, add markers and change clip speed from 0.25× to 4× with the pitch kept.

  • Edit right on the preview

    Click to select, drag with smart guides, scale and rotate. Double-click any text to edit it in place, with real line breaks.

  • Captions and text, Vietnamese included

    Word-highlight captions from your transcript — made on your PC, with ElevenLabs Scribe or imported — with rounded text boxes, stroke, shadow, highlight and emphasis colours, and reusable styles. Vietnamese diacritics render perfectly.

  • GPU-accelerated export

    A 71 s short exports in about 11 s on an NVIDIA GPU — 6.5× faster than real time. Other GPUs, and PCs without one, run the same engine with a software encoder.*

  • Built for AI agents from day one

    A built-in MCP server with more than 120 tools gives Claude, Codex, Gemini and other MCP agents the same controls you have. Several agents can build one video at once, every edit labelled with the agent that made it, with safety rails: rejected unsafe edits, no overwrites, and undo for everything.

  • Transcription & connectors

    Transcription on your PC. No key needed.

    Whisper runs on your computer, offline, in Vietnamese or English. Prefer ElevenLabs Scribe? Add your key in Connectors — the same page turns on AI voice, music, images and video, sent straight from your computer to the provider.

    1. Pick the voice track or any clip and choose Generate transcript.
    2. Transcribe on this PC — the speech model downloads once — or with ElevenLabs Scribe.
    3. Edit the transcript; captions update with it.

    Settings · Connectors

    1 of 3 keys added

    Connectors

    • ElevenLabs
      Scribe transcription · voice · music
    • OpenRouter
      Images and video
    • Vbee
      Vietnamese voice
    Transcript · voice.wavvia ElevenLabs Scribe

    00:02.60Stop saying “very good”.

    00:04.10Say excellent or brilliant instead.

    00:06.80Native speakers use it all the time.

  • Keyframes, camera and 35 transitions

    Keyframes with seven easings, zoom regions, a camera track that pans, zooms and rotates the scene while captions stay put, motion blur, and 35 transitions from push and flip to glitch and ripple.

  • An Elements library that moves

    71 animated elements: charts and counters, a Vietnam map, bar chart races, lower thirds, kinetic text, social-app mockups, audio visualizers, 3D titles, 60 animated emoji, plus Lottie and Rive imports — brand-aware and timed to spoken words.

  • AI voice, music, images and video

    Add your own key in Settings → Connectors to generate voice-overs (ElevenLabs, or Vbee for Vietnamese), music that ducks under the voice, and images or short videos through OpenRouter. Every new video asks you to confirm its price first.

  • Silence removal and scene detection

    Remove silences from a talking-head take after a quick review, split long recordings at scene changes, and let your agent apply a whole batch of edits as one undo step.

  • Brand kits, references and stickers

    Keep logos, colours, fonts, text styles and voice in a brand kit linked to your projects, study reference videos down to a shot-by-shot blueprint, and drop in bundled or your own animated stickers.

  • What you see is what you export

    Preview and export follow the same rendering and colour rules. Measured difference: under 1 % of pixels.

  • Export settings you’d expect

    480p to 4K (2K included), 24 to 60 fps (including 29.97 and 59.94), H.264 or H.265, MP4 or MOV, three bit-rate levels, and any output folder.

  • Local-first projects

    A project is a plain folder on your machine — a project.json and your media. No account, and nothing needs to be uploaded to edit.

* Measured on the developer’s machine with an RTX 5060 Ti. See Performance above.

Demo

Watch one short go from draft to export.

Four chapters: import, ask your agent, polish, export.

Pricing

Free to start. Pro when you’re ready.

Every editing and AI feature is in the free version. Pro removes the watermark and unlocks 4K and H.265 export.

Free

$0 forever

  • The full editor, captions and effects
  • AI agents via MCP — Claude, Codex, Gemini
  • Transcription on your PC, plus your own AI keys
  • GPU-accelerated export
  • CoClip watermark on exports
  • Export up to 1080p, H.264
Download free for Windows
Save $140 with yearly

Pro

$20 / month

or $100 / year — that’s $8.33 a month

  • Everything in Free
  • No watermark
  • Export up to 4K, H.264 or H.265
  • Use on up to 2 devices — Windows or Mac
Get ProCompare plans in detail →

FAQ

Questions, answered.

What do I need to run CoClip?

Windows 11 (64-bit) or macOS [ADD MINIMUM macOS VERSION]. On Windows, an NVIDIA GPU is optional — it speeds up export, and CoClip switches to a software encoder without one; the installer is per-user and doesn’t need admin rights. [ADD MINIMUM CPU / RAM]

Does it work offline?

Yes. Editing, export and on-device transcription run on your machine, so you can work without a connection; the speech model downloads once, on your first transcription. Talking to your AI agent and the optional AI connectors (ElevenLabs, OpenRouter, Vbee) need an internet connection.

Where are my files stored?

On your computer. Each project is a plain folder with a project.json file and your media. Nothing is uploaded to edit, and CoClip has no cloud or account of its own. On-device transcription never leaves your PC; when you use a connector such as ElevenLabs Scribe, only what you choose is sent, straight from your computer to that provider.

How does transcription work?

Choose Generate transcript on a clip or the voice track. By default Whisper runs on your PC, in Vietnamese or English, with a Best quality (574 MB) or Fast (264 MB) speech model that downloads once. You can also add an ElevenLabs key in Settings → Connectors and use Scribe instead; ElevenLabs bills your account directly, and your key is stored encrypted on your computer. The transcript lands on the timeline and generates word-highlight captions.

Which AI agents work with CoClip?

Claude, Codex, Gemini and other agents that support MCP. CoClip includes a built-in MCP server; you connect your agent with your own account, and it gets more than 120 tools to read, edit, preview and check the project. Claude can also run in a panel right inside the editor, and several agents can build one video at once. Every edit appears live in the editor, is labelled in the AI log and can be undone. See the MCP setup guide for steps.

Does it support Vietnamese?

Yes. Text and captions render Vietnamese diacritics correctly — Ở, Ữ and the rest. On-device transcription understands Vietnamese, numbers in elements use the Vietnamese format by default, and Vbee adds Vietnamese voice-overs.

Which export formats are available?

MP4 or MOV, H.264 or H.265, at 480p, 720p, 1080p, 2K or 4K, at 24, 25, 29.97, 30, 50, 59.94 or 60 fps, with Lower, Recommended or Higher bit rate.

Does CoClip work on a Mac?

Yes. The same editor, agent integration, transcription and project folders work on macOS as on Windows. NVIDIA-accelerated export is Windows-only. [CONFIRM APPLE SILICON / INTEL SUPPORT]

How much does CoClip cost?

CoClip is free, with a watermark on exports and export up to 1080p. Pro is $20 a month or $100 a year: no watermark, up to 4K and H.265, on up to 2 devices. AI agents and the optional AI connectors bill you separately through your own accounts. [ADD REFUND POLICY]

Stop fighting the timeline. Start describing the edit.

Download CoClip for Windows or Mac and finish your next short with your AI agent.

Download for WindowsRead the MCP setup guide