# AI video is becoming instant. Finishing is still the bottleneck nobody demos. — retained record

> Retained reddit record from reddit:generativeAI.

Canonical: https://highsignal.app/data/records/27238495433aae65

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:generativeAI)  published 2026-09-22 17:41 ingested 2026-09-23 02:42

[original source ↗](https://www.reddit.com/r/generativeAI/comments/1wngbb4/ai_video_is_becoming_instant_finishing_is_still/)

## retained text



```
This week I watched demos of video changing while someone prompted it live. It looked like the future.

Then I went back to a real project and spent an hour choosing between two opening shots, fixing captions, removing one awkward breath, reframing for vertical, and checking whether the music drowned the last sentence.

Generation keeps getting faster. Finishing remains a pile of small decisions.

That is not a complaint. Those decisions are where the video becomes specific. I use AI tools, including Cardboard and Opus Clips for some first passes, but I still do not want a model deciding the final half-second before a cut just because it is statistically smooth.

The demo is always “look what it made.” The job is “would you publish this with your name on it?”

Which finishing task still eats the most time for you?
```



document fetched 2026-09-23 00:40 · canonical https://reddit.com/r/generativeAI/comments/1wngbb4/ai_video_is_becoming_instant_finishing_is_still
