A laptop AI turns client voice notes into n8n workflows
Clients explain automations in long, self-correcting voice notes, and decoding them is unpaid work. A freelancer in Mexico open-sourced a tool that does it on his laptop and answers with an importable n8n workflow and a reply full of the right questions.
Published Oct 23, 2026
The Wow
What someone built
Jesus Manuel Garcia Torres released Ni lo escucho (MIT) on Oct 23, 2026. Whisper transcribes the note locally, Gemma 4 fills a strict spec, and plain code compiles the n8n workflow, so the model never invents nodes. Drafts take about a minute from text and two from audio on his 16 GB laptop.
- Made by
- Jesus Manuel Garcia Torres (gartox) (not us)
- Original repository
- Ni lo escucho on GitHub
- License
- MIT License (repository created Oct 23, 2026)
- What you need
- Node.js
- Ollama with gemma4:e4b-it-qat
- Whisper CLI
- n8n to import the result
- A laptop with about 16 GB of memory (the creator's setup)
- Install
- Install Ollama and Whisper, run `ollama pull gemma4:e4b-it-qat`, then `git clone https://github.com/gartox/ni-lo-escucho`, `cd ni-lo-escucho` and `npm start`. Open http://127.0.0.1:3000, drop in a voice note or paste text, and import the JSON it returns into n8n.
- A business you could build on it
- Voice-note-first automation studio for small businesses
Why it's remarkable
It runs entirely on a laptop, costs nothing per note, and turns missing details into questions instead of guesses. The creator also published where it failed (wrong branches, a TikTok request, asking clients for passwords) and how he fixed each one.
The Opportunity
The business it reveals
A voice-note-first automation studio: the client sends a voice note, you send back a working draft and a fixed quote the same hour. The slow, unpaid scoping that makes small jobs unprofitable shrinks to a review.
Who would pay
Owner-run shops, clinics and offices that already run on WhatsApp and buy small automations, and the freelancers and agencies who serve them.
The Flip
3 ways you could turn this into a business
Voice-note-first automation studio for small businesses
- Who pays
- Owner-run shops, clinics and offices that run on WhatsApp
- How it would earn
- Fixed-price small automations plus a monthly care plan (our hypothesis; price untested)
- Why it might work
- The creator already sells this work to these clients; the tool removes the unpaid listening and sketching that makes small jobs unprofitable.
- What already exists
- Freelancers on Upwork and Fiverr, Zapier and Make experts, and owners building their own with n8n templates. The edge is speed to a quote and accepting voice notes as the brief.
Spanish-first automation shop for Latin American small businesses
- Who pays
- Small businesses in Mexico and Latin America that buy automations in Spanish
- How it would earn
- Per-project builds priced for local markets (our hypothesis)
- Why it might work
- The tool, its replies and its tests are Spanish-first, and most automation tooling and freelancers market in English.
- What already exists
- Local web agencies and English-language automation freelancers; few sell small automations in Spanish from a voice note.
Voice-to-workflow intake for internal ops teams
- Who pays
- Operations leads at 20- to 200-person companies
- How it would earn
- A setup project and a retainer to run the request queue (our hypothesis)
- Why it might work
- Staff describe the automations they want in messages and meetings; a local intake that drafts the workflow and its questions shortens the queue without sending data out.
- What already exists
- Internal IT ticketing, Zapier's and n8n's own AI builders, and consultants. The local, data-stays-home angle is the difference.
Why we'd try "Voice-note-first automation studio for small businesses" first: It's the creator's own business with the slowest step removed: you answer a voice note with a working draft and the right questions before a competitor has replied.
Build It
How you could build it
An intake loop: voice note in, draft workflow and question list out, a fixed quote, then delivery and a monthly check.
First step: Install it, record yourself describing a simple automation the way a client would, and import the result into n8n to see how close the first draft lands.
Agents and tools
- Ni lo escucho
- Ollama with Gemma 4 E4B
- Whisper
- n8n
- An AI agent to adapt prompts and check the work (ChatGPT, Claude, Gemini, Grok Bot or similar)
Setup
- Install Ollama, Whisper and the app, and run the demo note.
- Record five realistic requests from your niche and check each draft by executing it in n8n.
- Write three fixed-price tiers for small automations.
- Offer existing contacts: send a voice note, get a draft and a quote within the hour.
- Track how much each draft needed changing.
What the agent does
- Transcribes the voice note
- Fills the workflow spec and lists missing data
- Compiles the importable n8n workflow
- Drafts the reply with questions for the client
What you do
- Reviews every draft and fixes the logic
- Connects accounts with the client
- Tests, prices and delivers
- Keeps the client relationship
Your first customer: A business owner who already sends you voice notes about what they want automated.
Test demand first: Offer voice-note quotes to ten past or warm contacts. Success: three send a note and one buys within two weeks. Failure: nobody sends one, which says the intake isn't the bottleneck.
Your starter instructions
Paste this into the AI agent you use (ChatGPT, Claude, Gemini, Grok Bot or another) as your first message. Copying it starts work in your own agent; it doesn't create or deploy a business by itself.
Running it
What stays human: Reviewing every draft, connecting the client's accounts with them, testing in n8n, pricing and selling the job, and answering the client's follow-ups.
- Startup cost (estimate)
- About $0 in software (estimate) (Estimate: the tool is MIT-licensed, and Ollama, Gemma, Whisper and self-hosted n8n are free to run; the creator uses a 16 GB laptop. Not measured)
- Running cost
- Near $0 per voice note (estimate) (Estimate: everything runs locally, so the creator reports no per-request fees; client automations carry their own n8n and app costs. Not measured)
- How much the agent runs
- People run it; AI helps with tasks.
Pricing your own version? Use the free startup-cost worksheet.
The evidence
What the source shows: The creator's DEV post and the MIT repository with its tests and published eval answers (Oct 23, 2026). We read both; we haven't run it.
What's ours: The studio, its pricing and the flips are our application.
Worth planning around
- His tests use 12 fictional requests; he calls the scores a direction, not a benchmark.
- From audio, branches were right in 6 of 8 requests.
- A commenter flags that accented field names may be mangled; test before relying on them.
- The repository is days old.
Sources
- Ni lo escucho on GitHub (MIT; README, tests and eval results; created Oct 23, 2026)
- Jesus Manuel Garcia Torres on DEV: My clients explain automations in rambling WhatsApp voice notes (Oct 23, 2026)
- Jesus Manuel Garcia Torres on DEV
Sources reopened Oct 23, 2026. Spot a mistake? Tell us and we'll correct it in public.
Keep reading
- How people make money with n8n (guide)
- How much of the work do AI agents actually run? (guide)
- He still can't write code. His five-person ad agency now runs on about 270 tools he built with Claude. Every owner-run firm needs its first five
- A marketer who doesn't code built an elk-call coaching app with Claude Code and says it hit a $650K run rate in 110 days. Every skill learned by sound is the same opening
- A free n8n workflow flags fake supplier invoices before anyone pays them — and Germany's e-invoice switch makes it a service bookkeepers can sell
- Marc Lou's agent chased a broken-shoe refund through three support desks and got his 850 AED back. Small stores now need a claims desk that can answer agents
- Every discovery and launch plan