How to Turn a Voice Note Into a LinkedIn Post Automatically (n8n + Claude)
The short answer
Record a voice note, and an n8n workflow transcribes it with a speech-to-text model, then sends the transcript to Claude with a prompt that rewrites it as a LinkedIn post in your style. The finished post lands in a draft (or posts automatically) so 30 seconds of talking becomes a publish-ready post. You can trigger it from Telegram, email, or a simple upload form.
Which tool should you use?
Use n8n if you want this fully hands-off and already run n8n. Most solo creators do not need it for this.
- switch Apple Shortcuts + ChatGPT, you’re solo and want the far simpler no-install version on your Mac or iPhone
Writing LinkedIn posts is a tax on busy people: the idea takes ten seconds, the writing takes thirty minutes. This workflow flips that. You ramble into your phone for half a minute, and a polished post comes back written the way you write. Here’s the build.
What this workflow does
Three steps, start to finish:
- You send a voice note (from Telegram, email, or an upload form).
- n8n transcribes the audio to text.
- Claude rewrites the transcript as a LinkedIn post in your voice, and delivers it where you want it.
No editing, no staring at a blank page. The idea in your head is the only input.
What you’ll need
- An n8n account (cloud or self-hosted).
- A transcription model, Whisper (via the OpenAI API) is the easy default.
- A Claude (Anthropic) API key for the writing step.
- A way to send yourself the audio. Telegram is the smoothest, but an email or a form works.
Build it step by step
Step 1, Catch the voice note
Add a Telegram Trigger node and connect it to a bot you create with Telegram’s BotFather (it takes two minutes). Now any voice note you send that bot starts the workflow. Prefer email? Use the Gmail Trigger filtered to a label like voice-post instead.
Step 2, Transcribe it
Add an OpenAI node set to the transcribe / Whisper operation, and pass it the audio file from the trigger. Run it once. You’ll see your spoken words come back as clean text. That text is now the input for the writing step.
Step 3, Let Claude write the post
Add the Anthropic (Claude) node, connect it, and give it a prompt like this:
You write LinkedIn posts in my voice. Below is a raw, spoken voice note.
Turn it into a single LinkedIn post: a strong one-line hook, short punchy
lines, one clear takeaway, no hashtags, no emojis unless they fit naturally.
Match the voice in these examples of my past posts:
--- [paste 2-3 of your real posts here] ---
Voice note transcript:
{{ $json.text }}The two or three example posts are what make it sound human and like you, don’t skip them.
Step 4, Deliver it
Add a final node to put the post where you’ll use it: a Gmail node to email it to yourself, a Google Docs node to append it to a running “drafts” doc, or a Notion node to save it to a content database. One tap to copy, one tap to publish.
Step 5, Turn it on
Flip the workflow to Active. From now on, every voice note you send becomes a finished post automatically.
Make it yours
- Multi-platform: add a second Claude node that reformats the same note as an X thread or an Instagram caption, one voice note, a week of content.
- A daily nudge: add a Schedule node that texts you “what did you learn today?” each morning so you actually feed it.
- Hook variations: ask Claude for three hook options instead of one, and pick your favourite.
This is the kind of automation that pays for itself the first week. Want it built for your exact tools and voice, or as a video you can follow along to? Drop it in the request queue.
Frequently asked
Does it post to LinkedIn automatically, or just draft it?
Either. LinkedIn’s API is restrictive for personal profiles, so most people have the workflow drop the finished post into a Google Doc, a Notion page, or an email draft for a one-tap copy-paste. If you post through a scheduler like Buffer, n8n can hand the post straight to it.
What do I use to transcribe the audio?
Any speech-to-text model works. OpenAI’s Whisper API is the common choice (a few cents per note), and there are dedicated n8n nodes for it. You send the audio file, you get text back, and that text flows into the next step.
Can I use ChatGPT instead of Claude?
Yes. Swap the Anthropic node for the OpenAI node and keep the same prompt. Claude tends to hold a personal writing voice well across longer posts, but both work, use whichever account you already have.
How do I make it sound like me and not like AI?
Paste two or three of your own past posts into the prompt as examples and tell Claude to match that voice. The model copies your rhythm, sentence length, and phrasing far better from examples than from instructions like "sound casual."
Want this one done for you?
Get the enriched version with the importable workflow file, or have me build and hand over the whole thing, tested, with a walkthrough.