YouTube's new AI is destroying your Pacing & Analytics | Creator News

Transcript

We’re starting today with an open-source escape route from the Adobe subscription trap. YouTube is testing an AI that fast-forwards viewers through your videos, destroying your pacing. And we’ll look at the theoretical exploits to bypass YouTube’s new courses attachment limits. Here are the creative views of the week. Let’s go!

A couple of weeks ago, we briefly talked about Adobe’s new Project Moonlight AI co-worker. As always, I advised you not to get locked into their expensive subscription ecosystem. There were questions, like from Manini, who asked for good alternatives to Adobe Photoshop. Photopea is a free web-based Photoshop alternative, but it’s not open-source. Generally, open source is the better long-term option.

My recommendation here is getting PhotoGIMP, a plugin that overhauls GIMP to mimic Photoshop’s UI, shortcuts, and layout. This significantly lowers the learning curve for those switching from Photoshop. However, if you’re used to Adobe’s error-fixing, you might need to relearn image manipulation from scratch. I’ve linked the GitHub repo in the description; feel free to try it out! If you have specific workflow questions, like Manini did, drop them in the comments, and I’ll address them when appropriate.

Moving on to YouTube, we have a feature that’s great for viewers but could be a nightmare for creators: Until April 27th, YouTube Premium users on Android and iOS can test a new experimental feature called Auto Speed on English videos. This adjusts playback speed automatically, speeding up during slow sections and slowing down when information density increases.

While this can be helpful for viewers, it also overrides the creator’s creative vision—pacing, dramatic pauses, comedic timing—all flattened by an algorithm optimizing for information density. My bigger concern is how this affects analytics. If viewers watch 50% of a video at 1.5x speed, average view duration (AVD) and user session time decrease. Does YouTube register this as a negative signal? If videos are frequently auto-sped, does YouTube internally flag them as boring or suboptimal? Without adjustments to these metrics, this feature could wreck the recommendation algorithm for slower, methodically paced content. Share your thoughts in the comments!

Staying on the platform but shifting to educational content, I recently got a notification about a new YouTube Studio feature: attaching files to videos in a YouTube Courses playlist. You can now attach up to five PDFs per video via the video elements tab. Why only PDFs? Because they must be hosted on Google Drive, whose malware scanners are designed for PDF structures. This doesn’t catch all threats, especially password-protected ones, but it saves development resources by preventing the upload of potentially shady executable files directly to YouTube.

As someone who frequently pushes platform limits (shout-out to the YouTube devs watching—love your work and always appreciate your feedback!), I can’t let this stand. Theoretically, you could bypass the PDF restriction using base 64 encoding. This converts binary data from a DaVinci Resolve preset or zip file into regular text, which you can then place in a PDF. After uploading to Google Drive and passing its scanner, viewers can download, decode, and retrieve the original file.

Creator News Update

Of course, this was just a thought experiment, and we certainly didn’t build an internal tool this week to test it (although it did work flawlessly). Before you get too excited, we’re not releasing that tool to the public, and I highly recommend against creating your own to share files on YouTube that exceed the limits outlined in the help article. Obfuscating files to bypass security scanners is a surefire way to land yourself in hot water with spam, scam, and deceptive practices policies, which is why I won’t be demonstrating it here.

Stick to standard PDFs for your courses, but honestly, YouTube, this needs to change. If someone with access to a basic LLM can bypass the restrictions within 30 minutes of trying, why are you artificially limiting uploads to PDFs only? Why not allow images, audio files, or project files for DaVinci or Lightroom presets?

If a malicious actor truly wants to distribute malware, they’ll either use the method we discussed or host it as a Google ad, which also lacks filters. This PDF-only rule is an arbitrary obstacle that doesn’t effectively stop scammers but hinders legitimate educational content creators from sharing functional files with their students. In my opinion, YouTube should lift this restriction.

Perhaps I’m being too optimistic and there’s an attack vector I’m missing, but what are your thoughts? Share them in the comments below.

B2B Update: YouTube Studio Media Kit

YouTube has updated the native media kit feature in YouTube Studio. Previously part of the Brand Connect feature set, it was separated out a few months ago, and now all YouTube partners can access and share their media kits with sponsors. This kit now includes data on income brackets and parental status within your audience—a nice addition, but likely not a deal-breaker in sponsorship negotiations.

In my experience, brands prioritize core demographics (age and gender) and the integration concept over these additional metrics. Whether 35% or 30% of your audience are parents is rarely decisive, but it does enhance your pitch deck’s appeal.

Auto Speak Feature: Helpful or Insulting?

As we wrap up this week, I want to know your thoughts on the auto-speak feature. Is it a valuable tool for viewers or an affront to video creators? Share your opinions in the comments below.

Stay tuned for more creator news next week. I’m Martin, signing off for now.