Instagram reel rewriter with an optional avatar video

Instagram reel rewriter with an optional avatar video

An n8n workflow that takes Instagram reel links from a Google Sheet and writes a new script, caption and on-screen line
for each, on the same topic and in the same shape as the original. An Apify actor transcribes the reel and OpenAI's
gpt-4.1-mini rewrites it. A second, switched-off part turns an approved script into a talking avatar video with
ElevenLabs and RunningHub, adds captions with Submagic and posts it to Instagram through Mixpost.

What is inside

• workflow.json: the workflow as n8n exports it, 41 nodes. Five are sticky notes that number its parts. The first part,
  eleven working nodes, is on. The other 25 working nodes are switched off, as they were in the original copy. It holds
  no credential.
• README.md: these notes.

What we changed from the original copy

• The code in fixURL had lost its backslashes in the original copy, so its pattern read
  /https://www.instagram.com/[^/]+/(reel|reels|p)//, which is not valid JavaScript, and every run failed there. It is
  back to /https:\/\/www\.instagram\.com\/[^\/]+\/(reel|reels|p)\//, which takes the account name out of a link such as
  instagram.com/<name>/reel/<id>/.
• The Authorization header of Download IG Reel read Bearer <token>. It now reads Bearer <YOUR_APIFY_TOKEN>.
• The author's sheet is gone. Get row, Update row in sheet and Update row in sheet1 ask for YOUR_SHEET_ID and use
  the tab Sheet1.
• getImage pointed at a photo in the author's Google Drive. It now asks for YOUR_IMAGE_FILE_ID.
• Webhook had a random path from the author's n8n. It is now reel-avatar-video.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An Apify account and its API token.
• An OpenAI API key that can use gpt-4.1-mini.
• A Google account, and a Google Sheet (below).
• Only for the switched-off part: a Google Drive photo of the person the avatar shows, and accounts with ElevenLabs,
  RunningHub, Submagic and Mixpost. Its Mixpost nodes come from the community package n8n-nodes-gravitysocial, which
  must be installed under Settings, Community nodes. Install it before you import, or n8n shows those three nodes as
  unknown.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Make the sheet below, then copy its ID: the long part of its address between /d/ and /edit.
3. Add a Google Sheets OAuth2 credential. In Get row and Update row in sheet, select it and replace YOUR_SHEET_ID
   with the ID.
4. In Download IG Reel, replace <YOUR_APIFY_TOKEN> in the Authorization header with your token. The brackets go too.
5. Add an OpenAI credential and select it in OpenAI Chat Model.
6. Fix the retry loop as Good to know shows.
7. Put one reel link in ReelURL on the sheet's second row and leave the other columns empty.
8. Save and press Execute workflow. When Script, Caption and Overlay fill in, press Publish.

The switched-off part

1. Install n8n-nodes-gravitysocial if you have not.
2. In API Setup, replace YOUR_API_KEY with your ElevenLabs key, YOUR_VOICE_ID with an ElevenLabs voice,
   YOUR_RUNNINGHUB_API with your RunningHub key and YOUR_SUBMAGIC_API with your Submagic key.
3. Add a Google Drive OAuth2 credential. Select it in getImage and Upload file, and put the photo's file ID in
   getImage in place of YOUR_IMAGE_FILE_ID.
4. Add a Mixpost credential and select it in Upload video, Get many accounts and Create a post. Replace
   YOUR_WORKSPACE_UUID in all three with your Mixpost workspace.
5. Select your Google Sheets credential in Update row in sheet1 and replace YOUR_SHEET_ID there.
6. Select the 25 switched-off nodes and turn them on (the D key), then save.

The sheet

One tab, Sheet1, with these headers in its first row: ReelURL, Script, Caption, Overlay, Approved and
Video Link. Put a reel link in ReelURL and leave the rest empty. The first part fills Script, Caption and
Overlay, and the second part fills Video Link. No node reads Approved, so use it as your own mark of which scripts
should become videos.

How a run goes

1. Schedule Trigger starts a run every minute. Get row reads the rows whose Script is empty, and Limit keeps the
   first one.
2. fixURL cleans the link, and Download IG Reel calls the Apify actor 3C7L8IMQOkq3isV2Y and waits for its reply:
   the reel's transcript, caption, hashtags and a status.
3. Check Status goes on when status is success. Any other answer goes to Wait, which pauses 2 seconds and calls
   the actor again (read Good to know before you publish).
4. Generate Script, Caption, Overlay sends the transcript, caption and hashtags to gpt-4.1-mini. It asks for a script
   of about 60 to 75 words in the same shape. It also wants a caption with up to five hashtags and an overlay line of at
   most eight words. Structured Output Parser holds the reply to Script, Caption and Overlay.
5. Update row in sheet writes the three into the same row, found by its row number.

The switched-off part starts at its own trigger, Webhook, so nothing in the first part calls it. Send it a POST, from
a button, a script or another workflow, with the row's new script and caption and its row number:

{
  "Script": "<the script from the sheet>",
  "Caption": "<the caption from the sheet>",
  "row": 2
}

1. API Setup holds the four keys. getImage downloads the avatar photo, and generateAudio has ElevenLabs read
   Script aloud with eleven_multilingual_v2.
2. uploadImage and uploadAudio send both files to RunningHub, and Merge2 with Aggregate puts the two replies in
   one list.
3. createTask starts RunningHub workflow 1996082549246083074 with the photo, the voice and a centered crop.
   checkStatus and Switch check it, after 6 seconds and then once a minute through Wait2, until it reads SUCCESS.
4. getOutput fetches the video's address, and Submagic Create Task sends it to Submagic with the Hormozi 2 caption
   template, zooms and B-roll on 60 percent of it. Check status and Switch1 poll Submagic through Wait6 until it
   reads completed.
5. Download Video fetches the result. Upload video gives it to Mixpost, Get many accounts finds the workspace's
   Instagram account, and Create a post makes a reel post with the caption.
6. Upload file is meant to save the video to the top of your Google Drive, and Update row in sheet1 to write its
   link into Video Link.

Change the look

• The script's rules: the message in Generate Script, Caption, Overlay. It sets the length, the hook, the shape, a
  closing call to follow and an example of the avatar's way of speaking. Every script ends with the same follow line,
  so change that line to your own.
• The caption and the overlay: steps 2 and 3 of the same message.
• How often it looks for new links: Schedule Trigger.
• The voice: voice_id in API Setup, and model_id in generateAudio.
• The captions' style: templateName, magicZooms, magicBrolls and magicBrollsPercentage in Submagic Create Task,
  and its fixed title, which reads My Awesome Video.

Good to know

• The retry loop: on a retry, Download IG Reel reads ReelURL from the item in front of it, which is then the actor's
  reply and not your row. Unless that reply has a ReelURL field, the call goes out with an empty address, fails again
  and repeats every 2 seconds, each time as a new paid actor run. Before you publish, change the address in its body to
  {{ $('fixURL').item.json.ReelURL }}, or point the false branch of Check Status at a stop.
• Overlapping runs: the actor can take more than a minute to transcribe a reel, and the row's Script stays empty until
  the end. The next minute's run then picks the same row and pays for it again. Set Schedule Trigger to a longer gap,
  such as 10 minutes.
• Executions: a trigger every minute is 1,440 runs a day, the empty ones included. Check your n8n plan's allowance.
• Model shutdown: OpenAI's deprecations page, read on 2 November 2026, lists GPT-4.1 mini for shutdown on 14 January 2027,
  with gpt-6-luna as its replacement.
• Cost of a reel: the actor is sian.agency's Instagram AI Transcript Extractor. Its free tier listed $0.028 for each
  transcribed reel plus $0.005 for each start, read on 2 November 2026. OpenAI bills one gpt-4.1-mini call a row.
• The switched-off part has three faults to fix before you rely on it. First, the file field of Upload file is an
  expression that returns the video itself, not a field name. Its input also comes from Create a post, which carries
  no file, so it fails as written. Connect it after Download Video with the field data. Second, Wait6 waits only
  1 second between Submagic checks. Third, a RunningHub or Submagic status the two Switch nodes do not name stops the run
  with an error.
• Someone else's reel: the prompt keeps the original's topic and structure and changes the words. Check Instagram's rules
  on reused content, and the creator's rights, before you post a remake.
• Your data: the reel's link goes to Apify and its words to OpenAI. In the second part, the script goes to ElevenLabs,
  the photo and the voice to RunningHub, and the video to Submagic, Mixpost and Google Drive.
• Keys: in the second part the four keys sit in API Setup, a plain node, so take them out before you share the
  workflow. The Apify token sits in a header of Download IG Reel.