See llms.txt for all machine-readable content.

Back to Templates

Create narrated faceless YouTube Shorts with OpenAI, FLUX, Kling and ElevenLabs

Created by

Created by: Andy || banandyy
Andy

Last update

Last update 3 days ago

Categories

Quick overview

Turn an approved script into a narrated vertical Short with FLUX images, selected Kling motion and timestamp-based captions. Review the exact storyboard and final video, reuse unchanged scenes, and archive a private Google Drive review ZIP.

How it works

  1. Reads a unique, rights-confirmed script from a Google Sheets Stories queue and freezes it in a versioned native Data Table record. Cost estimates, limits and an explicit generation switch are required.
  2. Uses OpenAI to map the unchanged script to three to six consecutive scene word spans and visual prompts. A signed-in, allowlisted n8n reviewer approves the exact storyboard hash before production continues.
  3. Requests ElevenLabs narration with character timestamps, parses the returned MP3 frames, enforces thirty to sixty seconds and validates spoken-text alignment.
  4. Caption cues and scene boundaries follow those returned times.
  5. Produces portrait FLUX 2 Pro images and selected five-second Kling 2.5 Turbo Pro motion clips through fal. Saved per-scene request IDs support bounded polling and reuse of unchanged assets.
  6. Uploads the verified narration through Shotstack Ingest and renders the measured timeline with timed rich-text captions. Original images remain visible after motion inserts end.
  7. The worker inspects actual PNG and MP4 bytes.
  8. Stores the final MP4, SRT captions, alignment and storyboard manifest in an owner-only Google Drive ZIP. Authenticated review checks the exact version and artifact hash; nothing is published to a social account.

Setup

  1. Import workflow.json, setup.json and review-and-handoff.json into one n8n project. Import demo.json to try the credential-free synthetic MP3/MP4 fixture first. Keep production inactive while configuring it.
  2. Connect Google Sheets and Drive OAuth in setup.json and run it once. Copy the returned spreadsheetId, stateTableId and private driveFolderId into both production and review settings.
  3. Bind OpenAI, fal, ElevenLabs and Shotstack credentials to the documented HTTP nodes. Use an authorized ElevenLabs voice ID and matching Shotstack stage or v1 keys. Enter current service estimates and a positive estimate limit; charges are separate from the template price.
  4. Replace the reviewer allowlist, retain signed-in Form authentication, and prepare an original approved sixty-five to one-hundred-sixty-word script with three to six scenes and explicit rights confirmations. Enable generation only when ready for paid use.
  5. Run one worker step at a time or enable the inactive minute schedule after testing. Inspect and approve the storyboard hash, then download, watch and listen to the exact final private ZIP before recording final approval. Follow README.txt for recovery and one-scene revisions.

Requirements

  • Current n8n with native Data Tables and signed-in Forms; Google Sheets and Drive OAuth; OpenAI API; fal access to FLUX 2 Pro and Kling 2.5 Turbo Pro; ElevenLabs with an authorized voice and timestamp endpoint; Shotstack Edit/Ingest. Separate provider and hosting charges apply. Uses built-in nodes only. Production generation is disabled by default. Local synthetic and provider-mock tests only; real providers and n8n Cloud have not been validated.

Customization

  • Change the approved script, audience, language, visual style and three-to-six scene count in Stories. Select zero to two motion scenes through the bounded storyboard configuration. Revise one visual scene with authenticated exact-version approval while retaining narration and unchanged assets. A script or voice change requires a new story ID. Other model adapters, music and social publishing require additional implementation and testing.

Additional info

The USD5 price covers the template package, not API credits or a hosted service. Final narration must measure thirty to sixty seconds; captions follow validated provider character alignment rather than an assumed speech speed. The local demo uses a test tone and artificial timestamps, not live generated speech. Container and hash checks do not replace full listening, visual, factual and rights review. Provider source URLs may be public or expire; the final archive is owner-only in Drive. Ambiguous paid POST responses stop for verified-ID reconciliation rather than automatic retries. No social platform upload or publication is performed.