Visgift Studio — Automated video pipeline

Context

Visgift ships short-form content (Nollywood / francophone) at a high cadence. Producing a video by hand — voice, captions, edit, publishing to every network — is slow and doesn’t scale. The goal: industrialize the chain end-to-end.

Technical challenge

• Chain a dozen heterogeneous tools (voice, edit, captions, CDN, scheduling) into one reliable, replayable pipeline.
• Generate karaoke captions synced to the word on synthetic voice.
• Publish automatically on 4 platforms with the right formats and metadata.

Solution

  Script ─▶ Voice (ElevenLabs / Chatterbox) ─▶ Karaoke captions (word-level)
                                                        │
                    Assets (images, music) ─▶ ffmpeg render ─▶ Vertical MP4
                                                        │
                             Cloudflare R2 (CDN) ─▶ Buffer ─▶ 4 platforms

• Voice synthesis (ElevenLabs + Chatterbox) driven by script.
• Karaoke captions generated at the word level and burned into the render.
• ffmpeg orchestrated from Python/Node for a reproducible vertical render.
• Storage / distribution via Cloudflare R2, scheduled publishing via Buffer on 4 networks.

Outcome

• 100+ shorts produced by the pipeline, at a steady cadence.
• Multi-platform publishing automated (×4), formats and metadata handled.
• Replayable chain: from a script to a published video with no manual step.

Learnings & trade-offs

• A media pipeline is tool plumbing before it’s creativity. The value is in the chain’s reliability, not in any single tool.
• Automated publishing forces upstream standardization of formats and metadata.