Pipeline Studio
Marketing pipeline automation

Run short-form video for every brand without touching a timeline.

Pipeline Studio is a three-lambda system: context, generation, render. It selects unplayed clips, feeds Claude and Runway/Pika with brand-safe prompts, normalises output with FFmpeg, and posts directly to Instagram Reels and YouTube Shorts—every single day.

Open pipeline console

Built on AWS Lambda • S3 • DynamoDB • SQS • EventBridge • Secrets Manager

Routes in rotation
MEME • ANIMATION
Daily pipelines
Per-brand, per-schedule
Status telemetry
Context → Render → Post
Pipeline telemetry
Next scheduled run
14:00 UTC · Brand: Moonshot Studios
Context lambdaQueued (SQS marketing-generate)
GenerationClaude caption + Runway pending
Render & publishAwaiting raw mp4
Instagram Graph API v19.0
YouTube Data API v3
Context Oracle
A Dynamo-backed brain that pairs fresh clips with brand personas, banned topics, and CTA rules before it ever hits AI.
  • • DynamoDB schema keyed by BRAND# + CLIP# records.
  • • Jobs stored in S3 jobs/{brand_id}/{job_id}.json for observability.
  • • Secrets + tokens resolved via AWS Secrets Manager.
Generation Studio
Claude captions and Runway animations orchestrated by SQS. No duplicate ideas, no overused hooks, no sleepy scripts.
  • • DynamoDB schema keyed by BRAND# + CLIP# records.
  • • Jobs stored in S3 jobs/{brand_id}/{job_id}.json for observability.
  • • Secrets + tokens resolved via AWS Secrets Manager.
Render & Publish Deck
FFmpeg normalises every frame from /raw, overlays safe captions, and hands off to Instagram Graph + YouTube Data APIs.
  • • DynamoDB schema keyed by BRAND# + CLIP# records.
  • • Jobs stored in S3 jobs/{brand_id}/{job_id}.json for observability.
  • • Secrets + tokens resolved via AWS Secrets Manager.

Every job is a contract between automation and your brand voice.

Pipelines are triggered by EventBridge, but brand managers can override schedules from the console. Each job logs state transitions—context → generation → render → publish—and the job JSON captures channel responses, media IDs, and error notes for audit trails.

Job payload excerpt
```json { "generation": { "provider": "runway", "caption": { "primary": "Deadlines chasing you?" }, "raw_s3": "s3://marketing-jobs/raw/post_2025_11_06_001.mp4" }, "publish": { "instagram": { "status": "POSTED", "media_id": "179568..." }, "youtube": { "status": "POSTED", "video_id": "YTs4T..." } } } ```
SQS queues
marketing-generate
Triggers content lambda (Claude / Runway).
marketing-render
Invokes render + publish lambda with FFmpeg.
Monitor job telemetry

Ship daily without a content bottleneck.

Brands supply clip libraries and guardrails; Pipeline Studio handles everything else. Each run picks the next clip in rotation, narrates it with fresh hooks, and ensures both channels stay in sync. Need a surprise drop? Queue it manually in two clicks.

S3 clips/{brand}S3 outputs/{job_id}.mp4Secrets Manager tokensRunway / Pika optional
Why it’s different
No editor
FFmpeg auto layout

Caption band uses a safe zone overlay; audio preserved with AAC normalisation.

API-level posting
No Zapier, no chrome bots

Instagram Graph container + publish flow, YouTube resumable upload with OAuth refresh.

Observability
Job JSON is the truth

Every update writes back to S3 so brand managers can download receipts instantly.

Extendable
LangGraph ready

Drop LangGraph inside the generation lambda for planning flows when you need deeper orchestration.

View architecture docs