n8n automation ideas lists skip video. Automate that step next.

Your inbox triage runs. Your CRM sync runs. Your Slack digest has been posting at 8am for months, and every "n8n automation ideas" list you open is still pitching those three. The question you actually have is narrower: which step in the pipeline you already built still stops and waits for a human.

Quick answer: The best n8n automation ideas for a builder whose text and data workflows already run are the media steps still done by hand: reframe a clip to 1:1 and 4:5, burn captions, stamp a watermark, change speed, cut shorts out of a long recording, and assemble several clips into one vertical MP4. n8n can orchestrate all six and execute none of them, because n8n Cloud has no shell and no Execute Command node, and n8n 3.0 turns unverified community nodes off by default. The hard way is self-hosting n8n in a Docker image with FFmpeg baked in and maintaining both; the easy way is one HTTP Request node to a video API such as FFmpeg Micro, billed on video length with a free tier.

Why video is missing from every n8n ideas list

Video is missing from those lists for a structural reason, not an editorial one: n8n can't run it. The general-purpose idea posts are built from triage, classifiers, digests and "content repurposers" that turn out to be a Basic LLM Chain rewriting text, because every item on them is work n8n itself can perform. Anything that needs a binary to touch a file gets quietly dropped.

The usual split when choosing what to build is deterministic work for n8n, reasoning work for an agent. That split is real, and it misses a third category: work n8n can orchestrate but not perform. Encoding, burning captions, cropping a frame and concatenating clips all live there. n8n Cloud gives you no shell access, no system packages and no Execute Command node, which is why the standard community answer to "how do I run FFmpeg here" is that you can't (n8n community).

Self-hosting used to be the escape hatch, and n8n 3.0 narrowed it. The 3.0 breaking changes flip N8N_UNVERIFIED_PACKAGES_ENABLED from true to false, so unverified npm community nodes stop installing and already-installed ones stop loading (n8n changelog). The 3.0.0 beta landed on October 27, 2026, with stable 3.x following in early December, so this is a this-quarter problem for anyone whose video step is a community package. The fix order is in n8n 3.0 disables community nodes: keep your FFmpeg step alive.

The shortage was never ideas. The public template gallery holds well over 11,700 community workflows. The shortage is steps n8n can actually execute, and one of the gallery's own entries proves it: workflow 9197 does voiceover and subtitles with Whisper, OpenAI TTS and FFmpeg, so it imports fine on Cloud and dies at the FFmpeg node.

Six video steps you're probably still doing by hand

These six are ordered by how often they're the last manual step in a working pipeline, based on what more than 300 builders have run through FFmpeg Micro from Make or n8n. Each one is a single job: one input file, one set of options, one output file. None needs a timeline or an editor.

The one-click versions of these jobs are in the blueprints catalog. Run one by hand on a real file, see the output, then build the node.

Reframing is the one with a deadline attached

Reframing is the highest-value of the six right now because the formats your ad placements want are not the formats your generator produces. The Gemini API exposes only 16:9 and 9:16 for every current Veo variant, and Sora's size options resolve to the same two shapes. That makes 1:1 and 4:5 a guaranteed post-generation step in any AI video pipeline, and it's a crop, not a re-render of the creative. The argument for reframing a master clip instead of regenerating variants is in Don't generate bulk video ad variants. Reframe your master clip.

Episodic shorts got a similar deadline on October 29, 2026, when YouTube added channel-membership monetization to Shorts Series and pitched it at creators writing multi-part dramas in 60-second scenes, citing that 40% of views and 30% of clicks arrive more than 30 days after publish (MediaPost). Consistent 60-second vertical episodes with per-episode burned captions is a production line, not an editing job. The twelve-originals version of that math is in YouTube Shorts bonus program: 12 originals from one long video.

Multi-clip assembly is what builders are actually asking for

Assembly is the hardest of the six and the one asked about most often in the n8n forums: several source clips at different frame sizes, a phone-recorded voiceover, hand-timed caption lines, all of it into one vertical MP4 (community thread). The trap is that concat refuses mismatched streams, so every clip needs normalizing to one resolution, pixel format and frame rate first, and then the narration has to sit under the whole thing rather than per segment.

The same job, as a command and as one node

The same comparison holds for all six. A 1:1 crop on your own machine is one line of FFmpeg:

ffmpeg -i master.mp4 \
  -vf "scale=1080:1080:force_original_aspect_ratio=increase,crop=1080:1080" \
  -c:a copy square.mp4

That line is correct and it's also the part n8n Cloud cannot run. Through the API, the same work is an HTTP Request node pointed at a URL:

curl -X POST https://api.ffmpeg-micro.com/v1/transcodes \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "inputs": [{"url": "https://storage.example.com/master.mp4"}],
    "outputFormat": "mp4"
  }'

In n8n the shape is four nodes, the same four for all six jobs: request a presigned upload URL, PUT the file, submit the job, then poll the job until it reports done and take the download URL. There are no completion callbacks, so a Wait node of a few seconds between checks is cheaper than a tight loop. Checking status, listing jobs and downloading are all free, so polling costs you nothing but executions. That's the bridge: if your video step is the one node you can't run where your workflow lives, FFmpeg Micro is that step as one HTTP Request, with the reframe, caption, watermark and speed options documented in the API docs.

Cost is per second of input video, not per render. One token buys 10 seconds of input, metered per second with no rounding up to the minute, so reframing a 30-second clip costs 3 tokens and producing all four aspect ratios from it costs 12. The free tier's 200 tokens cover about 33 minutes of input video with no card, which is roughly 60 reframes of a 30-second master. Trim a 30-second clip out of a one-hour recording and you're billed for 30 seconds, not the hour. Failed jobs are never billed.

Pitfalls that cost people a weekend

Four failures account for most of the broken video workflows people post about, and none are FFmpeg's fault.

When not to automate the video step at all

Automating a video step is the wrong call in two situations. If the output needs a designed layout, animated lower thirds, brand-locked typography, multiple moving text elements, you want a template-render service where a designer builds the template once; file operations expressed as API parameters are the wrong tool for layout work. And if you're processing hundreds of hours a day of long-form masters, owning the encoder starts paying for itself.

The input cap also sets a real boundary: 100 MB on the free tier, 1,024 MB on Starter at $19.99 a month, 2,048 MB on Pro at $49.99. A 4 GB camera master doesn't go through as-is, so that pipeline needs a proxy step first or stays on your own hardware.

FAQ

What should I automate with n8n after the basic workflows are running?

Automate the media step, since that's the one n8n can schedule but not perform. Specifically: reframing to 1:1 and 4:5, burning captions, watermarking, speed changes, cutting clips from long recordings, and assembling clips into one vertical MP4. Those six are where content pipelines still stop and wait for someone to open an editor.

Can n8n Cloud run FFmpeg?

n8n Cloud cannot run FFmpeg. It provides no shell access, no system packages and no Execute Command node, so there's nowhere to install the binary. You either self-host n8n in an image with FFmpeg baked in, or call a video API over HTTP from the managed instance you already have.

Will n8n 3.0 break the community node I was going to use for video?

n8n 3.0 sets N8N_UNVERIFIED_PACKAGES_ENABLED to false by default, so unverified npm community nodes stop installing and installed ones stop loading. If your video step is an unverified community package, it stops working on upgrade unless you set that variable back yourself. The ordered fix list is in n8n 3.0 breaking changes: fix your video step in this order.

Am I billed for render time or video length?

FFmpeg Micro bills on the duration of the video processed, not on how long the servers take. It's metered per second with no rounding up to the minute, one token per 10 seconds of input, and failed jobs are never billed.

Can I try a video step today without a card?

A new FFmpeg Micro account gets 200 tokens at sign-up, about 33 minutes of video, with API access, unlimited API keys and unlimited storage, and no credit card. What the free tier leaves out is customer support, token management, automatic top-up and token banking, which start at Starter.

Pick the step in your pipeline that still waits on a person, and build that one. The n8n Cloud walkthrough ships the workflow JSON for it, a main controller plus upload, metadata and wait-for-transcode sub-workflows, at How to use FFmpeg with n8n Cloud, and the 200 free tokens are enough to run the whole thing end to end before you decide it's worth keeping: sign up free.