🎬

NotTooOldForAI Video Pipeline

HeyGen avatar setup + article-to-video workflow. Enter hub password to continue.

Wrong password. Try again.

NotTooOldForAI

HeyGen Video Pipeline

One-time avatar setup + repeatable article-to-video workflow

← Back to Hub
The idea in one sentence: Record yourself once on HeyGen (2 minutes of webcam footage). After that, you never record again. Every article automatically becomes a 4-7 minute avatar video, which gets published to YouTube and cut into TikTok/Reels clips.

Why Instant Avatar, not Photo Avatar

HeyGen offers two avatar types. Use Instant Avatar (webcam video), not Photo Avatar (a still image). Here's why it matters for this specific use case:

Instant Avatar (webcam, 2 min)

Realistic motion for long videos. Holds up across 4-7 minute videos. Trained on your actual voice and natural movement. One recording session, used indefinitely.

Photo Avatar (still image)

Works fine up to ~15-60 seconds. Gets uncanny in longer videos: mouth movements become distracting and the stillness reads as "AI." Not suited for 4-7 min content.

Decision: Instant Avatar is the right choice here. The Annette videos are 4-7 minutes. Photo avatars fail at that length.

One-time setup (do once)

1

Create a HeyGen account

Go to heygen.com. The free tier gives a limited number of video minutes; the paid plan ($29/mo) gives unlimited short videos and avatar training. You need the paid plan for this pipeline.

2

Record your Instant Avatar (2 minutes)

HeyGen walks you through this with a script. You read naturally into your webcam for about 2 minutes. Good lighting, neutral background. This one recording trains your avatar forever. You never record again.

3

Wait for processing (24-48 hours)

HeyGen processes your footage into a trained avatar model. Once done, you can generate unlimited videos from text scripts using your face and voice.

4

Test with one article

Generate one 4-minute test video before building the full pipeline. Confirm the quality, voice sync, and output format work for you before committing 57 articles to it.

Repeatable article-to-video workflow

After avatar setup, this is the per-article workflow. Target: 1-2 long videos per week.

1

Convert article to spoken script

The article is the source material. Remove markdown headers, image placeholders, FAQ formatting, and CTAs. Rewrite any visual descriptions ("as you can see in the chart above") to verbal descriptions. Result: a natural speaking script, 600-900 words for a 4-7 minute video.

2

Paste script into HeyGen

Select your Instant Avatar, paste the script, choose background (plain light background works best for this style). Generate the video. Typical render time: 5-15 minutes.

3

Add captions + B-roll

Auto-captions are available in HeyGen or can be added in CapCut/DaVinci. B-roll = screenshots of the actual AI tool or prompt being discussed (screen recording, 5-10 seconds per concept). B-roll makes abstract concepts concrete.

4

Publish long-form to YouTube

Use the article's SEO metadata directly: the "SEO title tag" field becomes the YouTube title, the meta description becomes the YouTube description, the primary target query becomes the first tag. This reuses the GEO optimization already embedded in every article.

5

Cut Shorts for TikTok + Reels

Pull 3-4 clips of 30-60 seconds each: the hook (first 30 seconds), the quotable moment, and the best "aha" from the middle. These become TikTok posts and Instagram Reels. Each short gets its own caption (the article's quotable line works perfectly).

Output targets

  • Long videos: 1-2 per week, 4-7 minutes, YouTube
  • Shorts per video: 3-4 clips, 30-60 seconds, TikTok + Instagram Reels
  • Total short content per week: 3-8 short clips from 1-2 long videos
  • 57 articles = 57 potential long videos + 170-228 potential short clips
Next step: video scripts for the 57 articles haven't been generated yet. Once the HeyGen avatar is trained and tested, generate the spoken scripts from each article. This can be done in batches.

What's still not done

  • HeyGen account not yet created (Annette still needs to do the 2-min webcam recording)
  • 57 video scripts not yet generated (ready to do once avatar is trained and confirmed)
  • YouTube channel for NTOFA not yet confirmed as active