AI-powered content engine for videos and carousels.
Say it in plain English, post it everywhere.
II Content Engine turns plain-English prompts into publish-ready videos and carousels for TikTok, Instagram, and X. Open the repo in Claude Code or Codex, describe what you want, and the agent handles research, template selection, asset generation, captioning, and scheduled posting.
Everything is template-first — your ideas become reusable formats, not throwaway prompts. Generation is powered by Grok (API key or browser session), so you can produce content with or without an xAI API key. Approved assets flow into a scheduling queue that posts on your behalf.
Example output:
example/fruit-revenge-arc-api-portrait/fruit-revenge-arc-api-portrait.mp4
This is the current reference example for the video pipeline. It was generated from saved repo state, not a one-off prompt, and it follows the intended chain:
hero portrait -> reference sheet -> portrait scene-start frame -> clip -> stitched final
Example carousel:
- Preview:
example/fufu-how-to-carousel/preview.html - Research artifact:
example/fufu-how-to-carousel/research.json
- Videos — short-form viral clips, multi-clip story arcs, promos, character pieces, ASMR-style content
- Carousels — educational breakdowns, comparison lists, product showcases
- Auto-captioned assets — post captions plus burned-in on-video dialogue captions for video workflows
- Cost tracking — xAI spend logging, monthly budget warnings, and hard caps for API-backed generation and posting
- Scheduled posts — queue approved content for TikTok, Instagram, and X
npm install
pip install -r requirements.txt
npx playwright install chromium
cp .env.example .envOptionally add your XAI_API_KEY to .env, or rely on browser-based Grok session reuse.
If you use an xAI API key, you can also set a soft monthly budget and a hard monthly cap:
XAI_SPEND_BUDGET_USD=25
XAI_SPEND_CAP_USD=40Then open the repo in Claude Code or Codex and ask:
generate a video about a tired founder who clones himself to finish work
generate a carousel about how to season a cast iron pan
generate a 3-clip promo video for a local bakery
Or run the saved revenge-arc example directly:
BROWSER_OVERRIDE=false node code/cli/video.js "bullied orange serum revenge cafeteria drama" \
--template anthropomorphic-fruit-revenge-drama \
--md example/fruit-revenge-arc-api-portrait/fruit-revenge-arc-api-portrait.md- You describe what you want in plain English
- The agent researches the format and finds or creates a reusable template
- Grok generates images and video clips from the template's prompt contract
- The pipeline stitches clips, adds captions, and saves all artifacts
- You approve and the agent queues the asset for scheduled posting
For video work, the agent can scaffold a saved compilation markdown and asset-manifest.json from the selected template before rendering. That makes formats like fruit drama reusable without hardcoding fruit-specific logic into the renderer.
The repo now also has a higher-level video workflow CLI for prepare, render, queue, and ship, so the operational path is: define the reusable template/run artifacts, render the output, then move it into the scheduled posting queue.
For continuity-sensitive story videos, the default chain is:
hero portrait -> reference sheet -> portrait scene-start frame -> clip -> stitched final
There are two separate auth tracks:
Grok (generation) — works with XAI_API_KEY in .env or via browser cookies extracted from Chrome (npm run auth:grok). Generation works without an API key.
When XAI_API_KEY is used, billable xAI calls are tracked under output/tracking/. Run npm run spend:report to inspect the current totals.
Platform posting — requires saved cookies for each platform. Run npm run auth:posting (or platform-specific variants like npm run auth:posting:tiktok) to extract session cookies from Chrome.
See docs/INSTALLATION.md for the full setup guide.
For Dockerized posting/generation on hosts where docker compose build is missing buildx, use npm run docker:build. It builds both local images and falls back to direct docker build when needed.