Skip to main content

VidAI

VidAI

Record, understand and improve your videos — by talking to Claude.

A YouTuber records while VidAI robots run the camera, bring a new wig on request and paint a new background live

status: under development PyPI vidai-rec python 3.10+ Linux 61 tests passing

VidAI is a video plugin for Claude Code. Say vidai, answer a few questions, and record. While you record, VidAI measures small, useful statistics ("anchors") and applies live effects; you can talk to it ("VidAI, zoom in", "VidAI, add a title that says Hello"). Afterwards Claude uses the anchors to edit the video and exports a YouTube-ready file. Everything runs locally and for free.


How it works

VidAI flow: brief, anchor setup, record, live, anchors, understand, edit plan, render

Generated by assets/flowchart/flowchart.js (pure Node → SVG → PNG via headless Chrome): node assets/flowchart/flowchart.js


What it can do now

🎙️ Brief Claude asks what the video is about, the language, style, length and sections, then decides what to track
🎥 Record Its own recorder: camera, screen, or screen + camera, with mic · live preview · countdown · crash-safe files
📊 Live anchors Pauses, speech (speech-to-text), text on screen (OCR), motion, scene changes, active window, markers
🗣️ Voice “VidAI record / stop / mark / mistake / new section … / important / zoom in / zoom out / captions on” — anything else goes to Claude (English for now)
✨ Live effects Text, logos, arrows and boxes, zoom, blur, live captions — plus new effects Claude writes on the spot (e.g. an object that follows your hand)
💬 Talk or type Say “VidAI …”, hold ctrl+alt+space to talk without the wake word, or type in the Tell VidAI… box · “VidAI help” shows every command
🗣️ VidAI talks Say “VidAI talk” → “How can I help you?” → ask anything; Claude answers in VidAI's own voice (offline, natural), mixed cleanly into the video
↶ Undo / redo “VidAI undo” takes back a whole request (even one that changed several effects), “VidAI redo” brings it back
🙋 Claude asks you When a request is unclear, Claude asks in the window with buttons (“Which background? blur · purple · beach”) and tells you what it did — messages that are never recorded
🏎️ Stays smooth A performance governor keeps 30 fps: it slows tracking and preview first, warns you, and only then turns off the heaviest effect
🤔 Thinking icon A little animated VidAI eye bounces in the corner while Claude works on your request
🧠 Instant models Teach it something during the recording (“VidAI label yes”), correct it (“VidAI wrong”); it learns immediately
✂️ Editing Remove pauses and failed takes, add titles, arrows, zooms, subtitles, chapters, audio clean-up — without touching the original
🌱 Learns you Remembers misheard phrases you corrected, turns Claude's solutions into instant shortcuts, keeps your sizes and hand choices, and reads its lessons before every new video
🔐 Asks first “Master, I need to install … Say VidAI confirm” — or say “VidAI, take all actions” for full access
⚡ Fast Renders in parallel chunks split inside silences; hot loops in C
📺 YouTube export MP4 + subtitles (.srt) + a description with chapters

Quick start

pip install vidai-rec

Working on VidAI itself (tests, Claude Code skill and MCP config included):

git clone https://github.com/AlsammanAlsamman/VidAI && cd VidAI
python3 -m venv .venv && .venv/bin/pip install -e ".[all]"
.venv/bin/python -m pytest -q        # optional: run the tests

Then open Claude Code inside the folder and say vidai:

claude
> vidai

Claude interviews you, opens the VidAI Recorder, and edits the video when you are done. Prefer the terminal? vidai studio · vidai analyze video.mkv · vidai plan video.mkv remove_gaps · vidai render video.mkv

Requirements: Linux (X11), a webcam and mic (PipeWire/PulseAudio), gcc for the C speed-ups (optional). ffmpeg is bundled; the speech model downloads on first use.


Coming next

  • Speech-based editing: remove filler words, keep the best take automatically
  • Built-in tracking: hands, face, objects — effects that follow them
  • Arabic voice commands and on-screen text
  • More layouts, transitions and music

Metadata

Release files for vidai-rec 0.4.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for vidai-rec 0.4.0
File Size Uploaded
vidai_rec-0.4.0.tar.gz 770.2 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for vidai-rec 0.4.0
File Interpreter ABI Platform
vidai_rec-0.4.0-py3-none-any.whl Python 3 none any Details

Total release size: 1.5 MB

Release files / vidai_rec-0.4.0.tar.gz

Download URL vidai_rec-0.4.0.tar.gz
Size 770.2 kB
Tags Source
SHA-256 checksum
How to use checksums
b5f26b326d44083d612090324b5c92c7259e98d2d3e9e2e0ad90d250654fc3e2
BLAKE2b-256 checksum
How to use checksums
462c99632f678820c7dea4581c9137e0319593afa80e54b54263a94a8046f4b8
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.12.3

Release files / vidai_rec-0.4.0-py3-none-any.whl

Download URL vidai_rec-0.4.0-py3-none-any.whl
Size 758.2 kB
Tags Python 3
SHA-256 checksum
How to use checksums
1b94175d1ef4c8602fde10178dfa01985db756edf721a6d04176198d4bed9102
BLAKE2b-256 checksum
How to use checksums
90fab8bf63f8aaba7e588631901371a50d3432471dcf803c0b9e87c8ffc3e028
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.12.3

Release history Release notifications | RSS feed

This release

0.4.0 This release

2 release files

0.3.0

2 release files

0.2.1

2 release files

0.2.0

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page