What Matters in AI Shorts Production Is Not the Video — It's the Repeatable Routine

May 14, 2026

AIShortsNotebookLMGoogleVidsContentStrategyProductionRoutine

Summary

Running the first AI shorts production experiment with NotebookLM and Google Vids/Veo revealed that the core of shorts production is not video generation itself — it is building a repeatable routine that connects script, source, video, subtitles, upload, and review.

Fixing the Workflow Matters More Than Using More AI Tools

The biggest takeaway from this shorts production experiment was that fixing the workflow matters more than using more AI tools.

At first it looked like a single-video problem. There were individual decisions stacking up — what topic to pick, what kind of video to make, how to handle subtitles, which channel to post to. But working through the actual process made it clear that the core question was not about any one video. The real question was: can the same process be repeated the next time?

NotebookLM Is a SOP Accumulation Space, Not Just a Script Generator

In this experiment, NotebookLM was defined as a dedicated workspace for shorts production. Rather than simply asking it to generate a script, May operating standards, a shorts script template, a script generation prompt, a blog conversion prompt, and a performance log form were all loaded in as source material. Through that setup, NotebookLM moved beyond being a script generator and took on the role of a knowledge base for building a production SOP.

This distinction matters. AI content production does not sustain itself when you are always searching for a new tool. What matters more than the tool is the input standard, the script structure, the review criteria, the editing standard, and the post-upload logging method. Once those standards are in place, the next piece of content does not require starting from scratch.

Choose the Generation Method Based on Scene Information, Not Aspect Ratio

The video generation work with Google Vids/Veo also produced an important lesson. Generating directly in 9:16 seemed like the obvious approach for shorts, but the actual results told a different story. The 16:9 landscape video had more natural framing for the night-sea atmosphere and the headlamp inspection scenes. The 9:16 version fit the shorts format but came across as noticeably less realistic.

This changes the AI video production standard going forward. There is no reason to always generate in 9:16 from the start just because the output is a short. When scene realism matters, generating in 16:9 first and then cropping to portrait may produce better results. Conversely, when subtitle layout and fast turnaround take priority, 9:16 direct generation is the more efficient path.

Content Essence Determines Channel Selection

The haereujil lantern subject could have justified posting this as the first video on the ShoppingNotes channel. But the core of this content was not a product recommendation. It was an experiment — generating a script with NotebookLM, producing video with Google Vids/Veo, adding subtitles, and validating a shorts production routine. Posting it to Chulbuji Official as an “AI shorts production experiment” was the more coherent fit.

This judgment carries forward. A piece of content that covers a product is not automatically a shopping channel piece. The right question first is: is the essence of this content product information, an AI production experiment, or an operations retrospective? In this case the production method was what mattered, not the product.

Operating Standards from This Experiment

First, in May, validating the production routine takes priority over shorts monetization.

Second, NotebookLM is used not just for script generation but as a SOP accumulation tool.

Third, AI video generation is compared between the 16:9-generate-then-crop approach and 9:16 direct generation.

Fourth, even when shopping-related subjects are used, the initial format is a pre-purchase checklist, not a product recommendation.

Fifth, producing one video is not the end — the production process and decision criteria are recorded as a Log and added to the SOP.

The Core Insight

What matters in AI shorts production is not making one impressive video.

This experiment was not a process of making a polished short. Aspect ratio issues, realism gaps, subtitle typos, and channel selection questions kept coming up. But that is exactly what made the experiment valuable. Without actually running through it, the insight that 16:9 looks more natural, that subtitles in shorts need to be short, and that Chulbuji Official is a better first home than ShoppingNotes for this type of content — none of that would have been reachable.

What matters is building a routine: idea to script, script to video, video to edit, edit to upload, upload to log. Once that routine is in place, the question of whether to run a shopping shorts monetization channel in June can be answered with real data instead of assumption.