Part 7: Automate Video and Visual Content
Video outperforms text on every platform. But making videos is slow and expensive.
Unless your agent does it automatically. This lesson covers how to generate branded video and image content at scale — from dynamic per-article videos to static branded images — without a designer, editor, or manual rendering step.
Two Types of Visual Content
Jenny produces two distinct categories of visual content. Each serves a different purpose and uses a different pipeline.
Dynamic Videos (Per-Article, Auto-Generated)
Created automatically for every newsletter article. No human input needed — the pipeline picks up the article, generates a script, renders the video, and it's ready to publish.
Format
1080x1080 square format, 15–18 seconds long
Structure
Headline (2s) → 3–4 key points (3s each) → CTA (4s)
Visual Style
Brand colors, animated transitions, consistent look across every video
Pipeline
Claude writes the script → Remotion renders frame sequence → ffmpeg stitches into MP4
Cost
About $0.005 per video script via the AI API
Static Images (Branded, Platform-Specific)
Generated via Pillow (Python image library) for promotional and content marketing visuals. Each image is tailored to its destination platform.
Platform Dimensions
Instagram 4:5, LinkedIn 1.91:1, X 16:9
Brand System
Dark brown text, orange accent, warm white background
Features
Logo placement, dynamic font sizing based on text length
Use Cases
Tool promos, hub content promos, resource highlights
How Remotion Video Generation Works
The video pipeline is fully automated. Here's what happens from article to finished MP4:
Article Comes In from RSS Feed
The same content pipeline from earlier lessons picks up a new article. The video generation step hooks into that existing flow — no separate trigger needed.
Claude Generates a Video Script
The AI reads the article and produces a structured JSON script: a short punchy headline (max 8 words), 3–4 key takeaways with labels, and a fixed call to action.
Remotion Renders Each Frame
Remotion is a React-based video framework. It takes the JSON script, applies the branded template (colors, fonts, animations), and renders each frame as a PNG image at 2x resolution for crisp, high-quality output on any screen.
ffmpeg Stitches Frames into Video
The PNG frame sequence gets stitched into an H264 MP4 video at 30fps using ffmpeg. The result is a production-quality video file ready for any social platform.
Content-Hash Caching
Every video is cached by a hash of the article content. If the same article is processed again, the existing video is reused — no re-rendering. This prevents wasted compute and guarantees consistency.
Remotion is open-source and React-based.
If you can build a webpage, you can build a video template. The animations use spring physics for natural motion. The brand colors are baked into the template — every video that comes out looks like it was made by the same designer, because the template enforces it.
Pre-rendered Tool Promo Videos
For recurring promotions — tools, resources, assessments — Jenny uses a different strategy: pre-rendered videos with zero runtime cost.
6 pre-made videos for key tools
Prompt Library, AI Assessment, and other core resources each have a dedicated promo video ready to go at any time.
16.5 seconds each with a consistent structure
Intro (2.5s) → screen recording demo (10s) → CTA checklist (4s). Short enough for social, long enough to communicate value.
Stored as MP4 files, selected randomly
When the agent promotes a tool, it picks from the available promo videos at random. No rendering needed at runtime — instant selection and upload.
Zero runtime compute cost
Because these videos are made once and stored, they add no load to the server. The agent just selects a file and attaches it to the post.
When to pre-render vs. auto-generate
Pre-render when the content is stable and reusable (tool promos, resource highlights). Auto-generate when the content changes daily (newsletter articles, trending topics). Mixing both strategies gives you maximum coverage with minimum compute.
Visual Content Without Remotion
Remotion is powerful but requires infrastructure. If you want simpler options, there are solid alternatives:
Creatomate
API-based video generation. You design templates in their editor, then call the API with your data. No local rendering, no ffmpeg, no headless browser. Cloud-based and production-ready.
Shotstack
Similar to Creatomate — cloud-based video rendering via API. Define your edit as JSON, submit it, get back a rendered video URL. Good for teams that want to avoid infrastructure.
Canva API
Template-based image generation. If your team already uses Canva, you can programmatically fill branded templates with dynamic content and export images.
Plain images only
Skip video entirely and use branded static images. Pillow (Python), Jimp (Node.js), or even Canva templates can produce consistent, on-brand visuals without any video infrastructure.
Brand consistency matters more than video quality.
A well-branded static image beats a generic video every time. If you have to choose between a polished image with your brand colors, logo, and font — or a templated video that looks like everyone else's — pick the branded image.
Video generation requires the most infrastructure.
You need a Node.js runtime, a headless Chromium browser, and ffmpeg for encoding. If you're on a $5/month VPS, start with static images and add video later when you have the capacity. The image pipeline works everywhere — video needs more resources.
What You Just Learned
- Two visual content types: dynamic per-article videos and static branded images — each with its own pipeline
- The Remotion pipeline: AI writes script → React renders frames → ffmpeg makes video — fully automated
- Pre-rendered videos for recurring promos mean zero runtime cost — make once, reuse forever
- Start with branded static images, add video generation when you have the infrastructure
- Brand consistency matters more than production quality — a branded image beats a generic video
Next: You have content, distribution, control, and visuals. Now it's time to scale the entire system into a full marketing operation.