Content Studio — Video Generation (Synthesia)

Generate branded AI videos from your article using Synthesia templates. Avatar narration in 140+ languages with per-client template assignment.

What is it?

The Video mode in Content Studio → Visuals Studio. After an article is generated, pick a Synthesia template that's been assigned to the client, let Gemini auto-fill the template's variables from the article, and Synthesia renders an MP4 — typically a 30s–3min avatar narration. The generated video is saved to the draft so it reappears when you reopen it from the Asset Library.

Two prerequisites: a Synthesia API key (added in Data Sources) and at least one template assigned to this client (added in Agency Manager).

When should I use it?

A client needs a video version of a blog post for social — LinkedIn, YouTube Shorts, embedded on the article page.
You want a quick avatar talking-head video derived from the article copy.
You're A/B-testing video vs. text-only outputs and need consistent branding via your template.

How to use it

Connect Synthesia — Data Sources → Synthesia card → Verify API Key → Save & Connect. The app caches your full template list at save time. Status shows "N templates found and cached".
Assign templates per client — Agency Manager → click Edit on the client → scroll to Video Templates → click Sync from Synthesia if needed → tick the templates this client may use. Save.
Open the article — generate a new article OR open a saved draft from Asset Library.
Visuals Studio → Video mode — click the Video toggle in the Visuals Studio panel.
Generate Video — opens the picker modal.
Pick a template from the dropdown. Only templates assigned to this client appear.
AI auto-fills the text variables from the article in ~5 seconds. Edit any field manually, or click Regenerate AI auto-fill for a re-roll.
Leave Generate as test checked while you're iterating — test videos are free and watermarked. Uncheck to consume real quota.
Click Generate Video. The modal flips to "Generating…" — Synthesia typically takes 2–5 min for short templates, 5–20 min for multi-scene templates with several narration variables.
When done, the video plays inline. Use the download link or just leave it — it's saved on the draft and reloads automatically when you reopen.

Tips and best practices

Pick templates with fewer narration variables for faster renders. A 5-scene template can take 15+ minutes; a single-scene one finishes in ~2 min.
Use Test mode during template selection. Each Synthesia template has a different look and pacing — render 1–2 test videos per template before committing real credits.
The auto-fill quality scales with article quality. Short or vague articles produce thin narration. Generate the article first, review it, then generate the video.
Show template IDs to debug. Both the Manager picker and the modal display the template UUID so you can correlate against the Synthesia dashboard when something looks off.

Gotchas and limits

Express-2 avatars are NOT supported via the API. Synthesia's newer avatar generation is web-only. Templates using Express-2 fail at generation time with a clear error ("Pick another template, or recreate in Synthesia Studio using older 'Studio' avatars"). When creating new templates in Synthesia Studio, choose Studio-class avatars if you want API compatibility.
Templates with image or video variable slots fall back to template defaults. Only text variables are populated by the auto-fill (Phase 1 scope).
Test mode = free + watermarked. Production mode consumes minutes against your Synthesia plan. The cost-spike monitor catches anomalous spend daily.
Render generally takes 2–5 min, but multi-scene templates can take 20+ min. The Synthesia dashboard shows queue position. The poller checks every minute and auto-fails after 60 min of no progress on Synthesia's side.
Viewer role can see the rendered video but can't trigger new generations.