How to automate short video production: 8 n8n workflows from idea to publication

Why manual short-video editing eats up all your time

If you make short vertical videos regularly, you've probably run into the same thing: one video means a dozen manual operations. You need to write a script, record or generate voiceover, find suitable footage, edit it, burn in subtitles, upload to three platforms, and not forget about scheduling. At 5–10 videos a week, this turns into an assembly line that runs solely on your attention.

The problem isn't that you don't know how to edit. The problem is that the process isn't automated. Every step can be handed off to n8n — and then video production becomes a background task instead of a daily routine.

This article offers a practical solution: a set of 8 workflows that cover the entire cycle. You can build them yourself or take a ready-made foundation in the short-video pipeline kit, where all the steps are already connected to each other.

Overall architecture: 8 workflows and an orchestrator

The logic is this: one orchestrator launches the chain on a schedule, and each workflow handles its own stage. The separation matters because it makes debugging easier: if TTS breaks, you fix only TTS, not the entire pipeline.

Here's what the set looks like:

1. Orchestrator — scheduled launch and passing tasks between workflows.

2. Script generation — turning a topic or idea into the structure of a video.

3. TTS voiceover — voice synthesis and saving the audio.

4. Stock video selection — finding and downloading suitable clips.

5. Editing via FFmpeg — assembling tracks, trimming, overlaying sound.

6. Transcription via Whisper — getting timecodes for subtitles.

7. Burning in subtitles — rendering the final file with text.

8. Publishing — uploading to TikTok, Reels, and Shorts.

The orchestrator here isn't just a timer. It manages state: where the video is now, what's already done, what needs to be retried on error. This is a key element, without which automation turns into a fragile chain.

Step 1: Scheduled orchestrator

Start with the scheduler. In n8n, this can be a Cron trigger that runs, for example, once a day. The orchestrator reads the queue of topics from a database or Google Sheets and creates a task for a video.

It's important to build in idempotency right away: if a video is already in progress, a repeated run shouldn't create a duplicate. To do this, store the status of each task — "queued," "voiceover," "editing," "published." This will save you from chaos when several runs overlap.

Step 2: Script and TTS voiceover

The script can be generated via an LLM node: topic in, short text out with a hook, main body, and call to action. Keep the length suited to the format: 30–60 seconds of voiceover is optimal for vertical video.

Then TTS. Choose a voice for your niche and be sure to save the audio to a separate file with a clear name. At this stage, it's useful to get the duration right away — you'll need it during editing to trim the footage.

Step 3: Stock video selection

There are two approaches here. The first is keyword search from the script via a stock service API. The second is a pre-prepared library of clips broken down by topic. The second option is more reliable: you don't depend on API limits or search quality.

The workflow should download the selected clips and put them in the task folder. If there are several clips, preserve the order — it will come in handy during editing.

Step 4: Editing via FFmpeg

FFmpeg is the heart of the pipeline. Here you assemble the vertical format (usually 1080×1920), overlay the voiceover, trim clips to the audio duration, and add background music if needed.

Tip: don't try to do everything in a single command. Break editing into substeps — clip normalization, track assembly, sound overlay. That way it's easier to find the error if something goes wrong.

Step 5: Subtitles from Whisper

Whisper provides a transcript with timecodes. This is critical for subtitles: without timecodes, the text will "drift." Once you have the segments, generate a subtitle file (SRT or ASS) and burn it into the video via FFmpeg.

Burning in is better than manual overlay: subtitles are guaranteed to be visible on all platforms and don't depend on player settings. Keep in mind that vertical format requires a large font and safe margins from the edges — otherwise the text will be covered by TikTok or Reels interfaces.

Step 6: Publishing to TikTok, Reels, and Shorts

The final workflow uploads the finished file to three platforms. Here it's important to remember the differences: each network has its own requirements for description, hashtags, and cover. Build separate fields into the workflow for each platform so you don't publish the same text everywhere.

If you manage several accounts, add rotation to the orchestrator — that way you won't hit limits or overload your audience with identical videos.

How this connects to other automations

A short-video pipeline rarely lives in a vacuum. If you're an info producer, it makes sense to connect it to content production as a whole — for example, via an automatic content factory, where videos become part of an overall content strategy. And if you work as a micro-agency, add lead generation and reporting to this so clients see results in numbers.

Common mistakes when building a pipeline

The first is the lack of statuses. Without them, you won't know at which stage a video got stuck. The second is hardcoded parameters: duration, font, hashtags. Move them into variables, otherwise every edit will require digging into the code. The third is publishing without verification: add a step with a Telegram or Slack notification so you can see that the video went live.

And most importantly — don't try to automate everything at once. Launch one workflow, test it on one video, then add the next. That way you'll understand exactly where your bottleneck is.

Conclusion

Short-video automation isn't about replacing creativity, but about freeing up time. Eight workflows cover the entire cycle: from script and voiceover to publishing on three platforms. A scheduled orchestrator makes the process predictable, and splitting it into stages makes it debuggable.

If you don't want to build everything from scratch, check out the ready-made short-video pipeline on n8n — voiceover, editing, subtitles, and publishing are already implemented there. It's a good starting point for adapting the process to your niche.