Repurposing One Piece of Content into a Week of Posts with an AI‑Driven Workflow

Description and use

Repurposing one long‑form asset into a week of social‑media posts via a deterministic, version‑controlled pipeline that chunks, generates, validates, and schedules content locally.

Problem

Turning a single long‑form asset into seven daily social‑media posts requires a repeatable, deterministic pipeline that can be version‑controlled and inspected locally before external API calls.

Outcome

The workflow produces a week’s worth of social‑media posts from a single source, with deterministic, idempotent steps, local validation, and auditability, enabling scaling without hidden costs.

Workflow

• Extract source material (download audio or clone article)
• Normalize audio to 16 kHz mono WAV
• Chunk text into overlapping 1,200‑token windows with tiktoken
• Embed chunks with a 7–8 B quantized model (optional)
• Generate draft posts via LLM with strict JSON schema
• Validate drafts (schema, length, fact‑check, deduplication)
• Schedule and publish posts, recording platform IDs
• Maintain audit logs and prevent re‑posting

Evidence

Turning a single long‑form asset — such as a conference talk, a deep‑dive blog post, or a video tutorial — into seven daily social‑media posts is a repeatable engineering problem.

Source

All steps are scriptable, version‑controlled, and can be inspected locally before any external API is called.

Source

Because webhooks (or any async callback) are delivered at least once, the generator must be idempotent: include the chunk index in the output file name and skip already‑existing files.

Source

A tiny scheduler (cron, systemd timer, or GitHub Actions) picks the next unpublished draft in lexical order, calls the platform’s posting API, and marks the file as posted_timestamp.json.

Source