Thirty posts in one run: bulk processing in n8n and Make

Thirty posts in one run: bulk processing in n8n and Make

Thirty identical tasks means doing the same thing thirty times, and by the twentieth item the mistakes from monotony begin. Declining is galling: the job is large and repeatable, and a workflow with a list at its input solves it.

Thirty identical tasks means doing the same thing thirty times, and by the twentieth item the mistakes from monotony begin. Declining is galling: the job is large and repeatable. Below is a method that solves this entirely, and the honest limits of where it applies.

In short, if you are pressed for time

Thirty identical tasks means the same thing thirty times, and by the twentieth item the mistakes from monotony begin.

Declining is galling: the job is large and repeatable — exactly the kind that pays.

The fix is a workflow in n8n or Make: a table at the input, a loop over its rows, generation, and writing the result. The loop (Split In Batches / Loop Over Items in n8n, Iterator in Make) runs a section of the scheme for every row of the list.

Ready-made templates already exist: the automation templates section (more than 3,000 n8n workflows and Make scenarios) has starting points for posts from a table, content plans, product listings and images in bulk.

Launching a template takes an evening: download the JSON, import it, connect your own accounts, test on three rows, switch it on.

Videos in bulk work differently: in Cerebrum or through the connector, each video is a separate task, and that is stated honestly.

Bulk processing breaks on predictable things — the limits are named honestly so you are not disappointed.

The rule is the same as everywhere: everything is automated except the decision.

In money terms this is a move to another class of orders: a set of listings runs $245–730, against a median one-off job of about $60.

Thirty identical tasks means the same thing thirty times

The client sends a content plan for the month. Thirty subjects, each needing a carousel.

Or a list of forty products, each needing a listing. Or twenty items, each needing a description and three images.

The work is not difficult. It is identical, and that is the whole problem: repeating one sequence of actions thirty times.

A person does it for two days and by the twentieth item starts making mistakes from monotony. And declining a job like that is galling: it is large and repeatable.

There is a construction that solves this entirely.

How bulk processing is built

The scheme is always the same, in n8n and in Make:

A table or content plan (Google Sheets, Airtable, Notion) at the input. A loop over rows. Inside the loop: generation of text or an image through the models connected in the workflow. At the output: the result written back to the table, or published.

The loop has different names. In n8n it is Split In Batches (Loop Over Items in newer versions), in Make it is Iterator. The point is the same: a section of the scheme runs for each item of the list, taking the next one each time.

You build the scheme once, supply a list of thirty rows, and the loop runs it thirty times. One launch. Thirty results.

All of this is native to n8n and Make. Schedules (Schedule/Cron), webhooks, conditions, branching and waiting for approval (for example, by a Telegram message or an email) are built in and run on their own, without you.

What can be run as a list

Practically anything done identically.

Content to a plan. A list of subjects at the input — posts, captions or carousels at the output.

Product listings. A list of items — descriptions, SEO texts, images.

Variants for testing. One offer — six to nine creatives with different angles of approach.

Personalisation. A list of clients — a letter or a proposal for each.

Ready-made templates for bulk processing

You do not have to build from scratch. Below are real cards from the automation templates section. Each opens by its link: the description is visible to everyone, the file depends on your plan (see below). For details of what is inside, read the card's description; what follows is only what it states.

Texts and posts from a table

Content plan and publishing from a list

Carousels

Product listings and images in bulk

Videos from a table, an advanced template

How to launch a template

Five steps, the same for any template on the list.

Step 1. Download the file

On the template card, download the JSON. Regular templates are available on the Creator, Make, N8N and Full plans; advanced ones only on Full (or for a one-off $9, or for 1000 XP).

Step 2. Import it

In n8n: Workflows → Import from File, and choose the downloaded JSON.

In Make: create a scenario, open the scenario menu, choose Import Blueprint and point to the file.

Step 3. Connect your own accounts

Service keys, your own table, your own social accounts. A template carries no one else's credentials; you set yours.

Step 4. Test on three rows

Not on thirty. Make sure the substitution works and the results differ as they should.

Step 5. Switch it on

In n8n the Active toggle, in Make switching on the scenario schedule. From then on the workflow runs by itself.

Automation courses come with the Make plan (“Automations in Make”) and the N8N plan (“Automations in n8n”); both courses are open on Full. The choice between the platforms is covered in Make or n8n.

If you build your own workflow

The order that works first time.

Debug on one. Run the chain for a single row: thirty failed results are thirty paid model calls wasted, one failed result is one.

Find what changes. Usually one or two variables from the table: a subject, a product name, a photo link. The rest are constants.

Wrap it in a loop, run three, and only then the full list. And work out the cost before launch: thirty runs means thirty charges at your services.

Where bulk processing breaks

Honest limits, so you are not disappointed.

Uniform results. Thirty pieces from one scheme can come out indistinguishably alike. The reader will see one post multiplied thirty times.

How to fix it: build variety into the scheme itself. Different types of hook, different formats, alternation. The input list should differ not only in subject but in angle.

Quality at the edges. A scheme debugged on a typical row will stumble on an atypical one. In a list of thirty, two or three are always special.

How to fix it: a condition (an IF or Switch node in n8n, a Router in Make) — if a row does not fit the main scenario, it goes down another branch.

Checking. Thirty results have to be looked at. It is quick, but it is not zero — allow time.

Cost. Bulk processing multiplies not only the result but the spend. Do draft runs on cheaper models, final ones on the ones you need.

Where the human stays

The rule is the same as everywhere: everything is automated except the decision.

Put a pause with a wait for approval into the workflow: the result goes to you in Telegram or by email, and the chain continues only after your reply. This is how, for example, the "posting factory with approval" from the list above is built.

Sensible stopping points in bulk processing:

After the first three — to check that the scheme works on real data.

Before publishing — if the workflow ends with sending to social networks.

On atypical rows — if a branch has sent something down another path.

Videos in bulk: how it really works

Workflows cope with texts and images on their own. Videos are different, and it is better to say so plainly.

A quality video is assembled by Cerebrum (the chat on the platform) or by the AI SKILLS connector for Claude (Full plan). Each video there is a separate task: Cerebrum has no auto-launch or schedule. Paid steps come with an estimate, and you see the price before launch.

How to do a batch. In the connector you give Claude a list in one conversation: ten scenes or ten products. Claude goes through them one by one, and before each paid step shows the estimate and waits for your consent. This is not "one launch for thirty videos" but one conversation in which videos are made one after another. For a shop's advert you can attach photos of the products.

If you need genuine auto-launch from a table, see the advanced daily short video template above: it runs on the services you connect in n8n yourself.

The combination is this: regularity, collecting the input data and publishing are done by n8n or Make, and quality editing to a brief is done by Cerebrum or the connector. They are not connected to each other directly: you pass the finished video between them yourself.

What this changes in money

Here is the main practical consequence.

You can take large orders. Thirty listings and three are one workflow, with a different run time. Before, people turned such jobs down, or took them and suffered.

Price stops being linear in your time. The client counts by the number of units, you spend time on the scheme. The difference is yours.

You get an offer your competitors do not have. Not "I'll make listings" but "I'll close your whole catalogue in a week". For a seller with a hundred items that is a fundamentally different conversation.

Price range. From our bot's feed — 5,986 orders over four months: marketplace listings run $245–730 for a set, a system or pipeline $490–1,460, against a median one-off job of about $60.

A set and a pipeline are exactly what a workflow with a list at its input produces. Piece work stays on the bottom line.

And repeatability. The content plan arrives every month, the catalogue grows. A workflow built once serves the client continually — and that is no longer a one-off order but a stream. In the export, 12.9% of clients speak directly of a stream.

What else is on the platform

Cerebrum. A chat where the task is set in plain words ("assemble a video from this text"), the right Gen AI models are chosen automatically, the price is visible before launch, and payment comes from your balance as work is done. A paid launch only after your consent: aiskills.team/cerebrum.

Registration is free and opens three days of Basic access (no card needed).

Raw material for thirty posts. The express course Freelance 2.0, Basic plan:

Skills for a batch run

Both skills are advanced: Creator and above, or a one-off $9, or 1000 XP.

What distinguishes a batch from thirty separate runs. A quality filter inside the pipeline: thirty generated posts are useless if you have to read all thirty to choose eight.

What to take on the platform for this task

Where to start today

One action. Take your last order where you did several identical things and count: how much time went on the first unit and how much on all the rest.

If the second figure is many times the first, that is exactly the volume a loop over a table removes.

Open the templates section, find the "captions from a table" starting point and run it on three rows.

And from that moment an order for thirty items stops being something you turn down.

What to read next

Build once, it runs forever — the foundation this builds on.

Content gets made while you sleep — the next step of automation.

Relaunch on a marketplace — where bulk processing is needed most.

What from your builds actually sells — what volumes you can take on.

AI agents: what they are and what they pay — a full guide to agents, with market rates.

See also on AI SKILLS