Auto Captions / Transcriptions Consistently Out of Sync

Hey, sorry for the late responses. I completely forgot about this, here’s more information.

OS: MacOS Tahoe 26.5
System: Macbook Pro Apple M3 Pro

On the Premiere Versions:

26.2.2:
On this one it’s the worst, I literally have to move each caption seperately to get them in sync. I deleted this version.
26.2 (Which I’m currently using): This one just has a slight out of sync issue. Whereby I can drag all of my captions 3-5 frames behind. It’s always a 3, 4 or 5 frame delay, Meaning auto-captions come too late. 3-5 frames late. Delay. I’ll attach a screenshot here, where you can see the original captions created by Premiere Pro, vs My final captions the ones that are “Upgraded to Graphic” The delay is around 3-5 frames.

 Other versions:
26.0 to 26.0.2: These versions and the versions before that work perfectly, zero issues when it comes to auto captions. Only came in the newer updates, I haven’t downgraded to 26.0.2 because these newer updates had solved my issues with “Scaling with Bezier, for zoom out and zoom in effects”, in versions 26.0 to 26.0.2 there was an issue where changing the Scale property would change the keyframe from “Bezier scaling” to Linear. So that was fixed, so I have to use these current versions for that to work. But this caption being out of sync is very irritating. Takes way more time in workflow.

Answers to you,

This is an edited timeline with cuts of some type? Is the transcription(s) of the source clips or a static transcription of the sequence?

I’ll put the information of my workflow basically in here, so you can understand what I’m doing.

I do short-form videos from Camera Footage,
I create a 1920x1080 60fps Sequence, 

Here I import a clip that has been filmed with a camera, I do cuts, I cut out empty spaces, put an Optical Flow on the clip to make it 60fps. Upscale it in Topaz video AI to make it 4k.
Then I take the clip, make a 1080x1920 60fps sequence, render that upscaled video’s audio as mp3, put it in Adobe Enhance, after that I import that, replace the original audio with the enhanced one. Transcribe the sequence, remove commas and periods and create a caption track with these settings:
 

And that’s basically it. Have done this for 3-4 years now, no issues. Only in the newer versions (26.2 and 26.2.2) the captions started going out of sync after they’re created. Never had an issue with this before, for years I’ve used the same workflow.

When you play the edited timeline and watch the transcript, is it exactly on time? Particularly for the parts that later show captions out of sync?

Good question, so I just analyzed it. The “Transcript” window itself that transcribed the sequnece before it’s even created to captions, is ALSO out of sync. And it’s visible and pretty easy to notice. The transcript follows but each word is highlighted on a 3, 4 or 5 frame delay.

So the transcript is already out of sync, then the captions generated from that transcript are obviously out of sync as well.

There’s also a weird behaviour that I wanna add here, the words that start with the letter S so like: “Settle” “So” “Suing” “System” the delay is very noticeable and way bigger.