⭐ Storyboard MCP · Flagship sibling · Video translation

David, in two languagesOne English speaker, two non-Anglo target languages — French and Chinese. Same chassis as Camille; flipped source language to prove the pipeline isn't biased.

A short career reflection in English, localized to French and Chinese via heygen-translate-speed. Same model, same call shape, same voice clone — only the source and targets change. Two languages instead of three keeps the demo focused; the cost scales linearly.

source
English · 30s · MP4
targets
FR · ZH
model
heygen-translate-speed
wall clock
~3 min (parallel)
cost
~$3.00

Source — English, the original

🇺🇸 Source · EnglishDavid, software-engineering origin story

Two translations — same voice, different language families

🇫🇷 French96s · $1.50
🇨🇳 Chinese96s · $1.50
Same speaker, two language families. French and Chinese share almost no phonemes — but the voice you hear in both is David's. The voice-clone model doesn't need a per-target reference; it extracts a single embedding from the source and applies it across all targets.

Pipeline trace — identical to Camille's, swapped source + targets

Stage 1 — SOURCE Source: raw.githubusercontent.com/livepeer/storyboard/.../david-source-en.mp4 Any public URL works. For YouTube source, use yt-dlp first.
Stage 2 — TRANSLATE (parallel × 2) For each target in ["French", "Chinese"]: create_media({action:"animate", model_override:"heygen-translate-speed", params:{video_url:<source>, output_language:<target>}}) → 2 MCP jobs, run concurrently → both done in ~3 min wall-clock
Stage 3 — DOWNLOAD curl -o public/chapters/multilingual-demo/david-fr.mp4 <fal-output-url> curl -o public/chapters/multilingual-demo/david-zh.mp4 <fal-output-url>
Output → 2 MP4s, 1080×1080 each, ~1.5 MB · same speaker, two languages

QC pass — what to look for in a video translation

TargetVoice matchLip-syncRegisterProper nounsDuration drift
French✓ Same voice✓ CleanConversational"David" preserved±0s · matches source
Chinese✓ Same voice✓ CleanConversational (Mandarin)未本地化（保留 "David"）±0s · matches source

Cost — honest end-to-end

Source hosting
$0.00
push to /public on Vercel
FR translation
$1.50
heygen-translate-speed · 30s
ZH translation
$1.50
heygen-translate-speed · 30s

Total: ~$3.00 for a 30-second clip in two non-source language families. Cost scales linearly with target count: 5 languages would be ~$7.50.

Why this sibling shows what the flagship can't

Camille goes French → English (close), Spanish (closer), Chinese (far). David goes English → French (close) and Chinese (far). The pair proves the chassis isn't biased toward a particular source language family or accent. Same code path. Same cost-per-call. Same QC checklist.

Open the playbook → Flagship: Camille (FR → 3 langs) → All showcases