n8n automation

AI songs for YouTube, automated from idea to upload

An n8n workflow takes an idea from the queue, writes the lyrics, generates a song with vocals, cover art and a video, then sends you a preview on Telegram. Nothing reaches YouTube until you approve it.

OpenRouter · fal.ai · YouTube Data API · Telegram

In short

Automated AI songs for YouTube is an n8n workflow that takes a track from idea to upload: it pulls a topic from a queue in a sheet, writes the lyrics, title, description and tags, generates a song with vocals, cover art and a video, and finally sends you a preview on Telegram. The video goes to YouTube only once you approve it, and to the channel you choose; our version handles four channels at once. It’s built for creators and companies running music channels who want to publish regularly without assembling every release by hand across several tools. You set the direction and guard the quality, and the automation handles the repetitive production.

Workflow

This is the workflow in n8n

A simplified, interactive version of our workflow. Drag the nodes and run it to watch the data flow.

Drag the nodes · click “Run”

Execution log

    AI songs → YouTube: Idea queue → AI writes lyrics and metadata → song, cover and video via fal.ai → Telegram approval → upload to one of your YouTube channels

    Screenshot from the build

    The full workflow in our n8n

    A real, unedited screenshot from the n8n editor. The interactive version above is simplified; here you can see every node, branch and error path.

    Full size
    n8n workflow: AI songs for YouTube
    Workflow in n8n. Built by Bartłomiej Chwajoł.

    Sound familiar?

    What this solves

    One track means several tools, downloading files and pasting descriptions around.

    The workflow joins every stage into a single run: lyrics and metadata from a language model, music and video from fal.ai, cover art from an image model and the upload through the YouTube API. Files move between steps on their own.

    AI can come up with the same title twice, or something you’d never publish.

    Before any music is generated, the workflow checks the title against a table of past releases and asks the model for a new one if it’s a repeat. You watch every video in Telegram and decide yourself whether it goes ahead.

    Music and video generation takes time, and external APIs sometimes fail.

    Jobs go to fal.ai through a queue: the workflow waits, checks the status and fetches the result when it’s ready. Retries are capped, and every failure arrives as a Telegram alert instead of vanishing quietly.

    How it works

    Step by step

    1. 01

      Start and configuration

      The workflow runs hourly or on demand. A CONFIG node keeps the settings in one place (style, channels, retry limits), and a Router sends the run down the right path.

    2. 02

      An idea from the queue

      The next idea comes from the sheet: topic, mood, genre and target channel. If the queue is empty, the run ends without calling any paid API.

    3. 03

      Lyrics, title, description and tags

      A model via OpenRouter writes the lyrics and metadata to a fixed JSON schema. The title is checked against a table of published tracks, and on a repeat the model tries again.

    4. 04

      A song with vocals

      The lyrics and style brief go to the MiniMax Music model on fal.ai. The workflow submits the job, waits, polls the status and downloads the finished audio, stopping with an alert if it runs out of retries.

    5. 05

      Cover art and video

      An image model via OpenRouter creates the cover, and fal.ai renders a video from it and the track. The same queue-and-poll pattern applies here.

    6. 06

      Preview and approval in Telegram

      The video arrives in Telegram with its title, description and buttons. A rejected release is logged and never appears on YouTube.

    7. 07

      Publishing and logging

      Once approved, a Switch picks one of the channels and the workflow uploads the video through the YouTube Data API. The link is saved to the sheet and you get a confirmation in Telegram.

    Implementation details

    A music channel lives on consistency, and consistency eats time: lyrics, recording, cover, render, description, tags, upload. This workflow strings those stages into a single n8n run, but leaves you the part that matters most: deciding whether a track should come out at all.

    A queue, not a lottery. You keep ideas in a sheet: topic, mood, genre, channel. The workflow doesn’t invent your channel’s direction; it works through your list. A CONFIG node holds every setting, so changing the style, the model or the retry limit doesn’t mean editing a dozen nodes.

    Long jobs without hanging. Music and video generation is asynchronous: fal.ai accepts a job and returns the result some time later. The workflow submits the request, waits, polls for the status and only downloads the file once it’s ready. Every loop has a retry cap, and when it’s hit the run ends with a message instead of hanging for hours. A separate Error Trigger posts every failed execution to Telegram.

    A human before publishing. Nothing reaches YouTube without your tap. In Telegram you watch the finished video with its title and description, then approve or reject it. Rejected releases stay in the log, so there’s a record of what the model shouldn’t repeat.

    Platform rules and rights. YouTube has policies on mass-produced, repetitive content and on labelling realistic AI-generated material, and music models come with their own licence terms. During setup we go through the terms of the models you pick with you and set up descriptions so they fit your channel’s rules from day one. Channel strategy and monetisation decisions stay with you; the automation takes the production off your plate.

    If you also want to publish short-form clips, read how to edit short-form video that sells. For animation or channel branding, see our motion design and video service, and for the basics of the tool, what n8n is.

    01

    Integrations

    • n8n
    • OpenRouter
    • fal.ai
    • MiniMax Music
    • YouTube Data API
    • Telegram Bot API
    • Google Sheets
    • n8n Data Tables
    02

    What you need

    • One or more YouTube channels and a Google Cloud project with the YouTube Data API and OAuth access to each channel
    • A fal.ai account with an API key for music and video generation
    • An OpenRouter key for the text model and the image model
    • A Telegram bot and a chat to receive previews and alerts
    • A sheet with the idea queue and a short style brief for each channel
    • An n8n instance with enough memory and disk space for audio and video files (ideally self-hosted)
    03

    How we tailor it

    • A different number of channels, each with its own style, model or language
    • Instrumental tracks instead of vocals, or a different music model
    • Vertical cuts of the same track for YouTube Shorts
    • Approval by several people, or on Slack instead of Telegram
    • Uploading videos as scheduled instead of publishing them straight away

    FAQ

    Frequently asked questions

    Does YouTube allow AI-generated music?

    Yes, within its rules. YouTube requires you to label realistic content made or altered with AI, and the YouTube Partner Program checks whether content is mass-produced or repetitive. That’s why the workflow relies on a varied idea queue and manual approval of every video. Always check the current rules in the YouTube Help Center.

    Who owns the rights to the generated songs?

    That depends on the terms of the model and platform you use, and on the law where you are. Before setup we go through the terms of fal.ai and your chosen model with you, but that’s no substitute for legal advice if you plan, say, to distribute through streaming services.

    Can I switch the music or video model?

    Yes. We call the models through the fal.ai and OpenRouter APIs, and their settings live in the CONFIG node. Switching models usually means a new model ID and adjusted parameters, with no rebuild of the rest of the workflow.

    What happens if song generation fails?

    The workflow keeps polling up to a set limit, then stops the run and sends an alert to Telegram. The idea can stay in the queue for another attempt. Nothing gets published half-finished.

    Does the video go live as soon as I approve it?

    In our version, yes: once approved, the video goes to the chosen channel and the link is saved to the sheet. You can also upload videos as private or scheduled and have them go live at set times.

    Get a quote

    Want this in your business? Tell us what it should do.

    Describe your process and the tools you use. We reply within one business day with an implementation plan.

    What do you need?

    Goal, deadline, links to inspiration — the more specifics, the more accurate the quote.

    Other builds

    More n8n automations

    1. 01 Allegro listing optimisation with AI Allegro REST API · OpenAI · Google Sheets · Gmail Every hour, an n8n workflow takes the next offers from a sheet, pulls their data and category parameters from the Allegro API, and AI writes a new title and description. Changes reach Allegro only after validation, and every offer is backed up first.
    2. 02 AI email assistant on Telegram IMAP · OpenAI · Telegram Bot API · SMTP An n8n workflow reads your inbox, filters out spam and sends the emails that matter to Telegram. You record a voice note with your answer, AI turns it into a finished email, and one button sends, revises or discards it.
    3. 03 AI-personalised email & SMS n8n · OpenAI / OpenRouter · Google Sheets · SMS gateway An n8n workflow pulls contacts from a Google Sheet, writes each person their own message, sends the email and, for mobile numbers, an SMS too. The sheet shows you straight away what went out and what was delivered.
    4. 04 Multichannel product listing WooCommerce REST API · Allegro API · Erli API · OpenAI You enter the product data once, and an n8n workflow picks the category on each platform, writes the copy, prepares the images and creates the product in WooCommerce plus offers on Allegro and Erli. At the end you get one record with the result from every channel.
    5. 05 AI WooCommerce product translation WooCommerce · WordPress REST API · OpenAI · Google Sheets An n8n workflow takes untranslated products from a queue, and AI translates their titles, descriptions, attributes and SEO meta into the languages you choose. Each result is checked for completeness and length before it reaches your store.
    6. 06 AI SEO landing pages for WordPress Google Search Console · OpenRouter · WordPress REST API · Google Sheets An n8n workflow takes a keyword from a sheet, analyses the search results, plans the structure and writes a landing page with links to your existing pages. The finished page lands in WordPress as a draft for you to approve.