Publishub launches AI copilot to predict viewer drop-off in short videos

Predicting viewer behavior

Publishub is a new AI editing copilot designed specifically for creators of short-form content like TikToks, Reels, and YouTube Shorts. The tool aims to help users see exactly where their audience is likely to stop watching before the video is actually posted or delivered to a client.

The platform works by watching a video file second by second to transcribe speech, measure pauses, and detect scene changes. It then generates specific notes that point to the exact second a viewer might lose interest, rather than offering generic advice like "improve your hook."

As getpublishub.com reported, the tool provides a reason for every note, such as an intro running too long or a shot remaining still for too long. This allows creators to learn patterns over time and provide evidence-based reasons for their editing choices to their clients.

Workflow integration

The tool is designed to work alongside a creator's existing workflow rather than replacing their current software. It does not require any plugins to be installed or a new editor to be learned, meaning it can be used with any app that exports standard video formats.

Users export their video from apps like CapCut, Premiere Pro, DaVinci Resolve, or InShot and upload the file to the platform. The system processes the file and provides a fix plan, which the creator then takes back to their original editing app to implement.

Because the tool does not connect to the editor or touch project files, the creator maintains full control over the final output. Nothing is cut or changed without the user's approval, and the original file remains untouched during the analysis process.

Verification and accuracy

One of the unique features of the platform is its commitment to verifiable results. The tool predicts how much retention should rise after a creator re-records a specific line, allowing the user to check the actual results against the prediction once the video is live.

Creators can upload a retention screenshot from Instagram Insights to allow the AI to read their real curve. This enables the platform to build an accuracy score over time, helping the creator learn what specifically works for their unique audience rather than relying on general trends.

For example, if the tool predicts a 65% retention rate for a new hook and the actual result is 68%, the creator can confirm the prediction's accuracy. This feedback loop helps creators refine their content strategy based on real data rather than guesswork.

Privacy and pricing

The platform is built by Valmir, a solo founder and UGC creator who is sharing the development process publicly. To ensure privacy, uploaded videos are kept in private storage and are only used to generate the specific analysis for the user.

The service uses Google Gemini to process the audio and frames, but the developer states that videos are never sold and are never used to train the company's own models. Deleting an account results in the immediate removal of all videos, screenshots, and analyses from the server.

Publishub offers a free tier where users can test one video without an account to see drop-off points. Paid subscriptions are available at $12 per month for the Creator plan and $24 per month for the Pro plan, both of which include full action plans and rewritten hooks.

Technical requirements

To use the tool, creators need to provide a video file in MP4, MOV, or WEBM format. The platform currently supports files up to 500 MB in size, making it suitable for high-quality short-form exports.

The analysis process is designed to be fast, typically delivering a fix plan in about 3 minutes. This speed is intended to keep the tool useful for fast-paced production environments where editors need quick feedback before moving on to the next project.

The tool is positioned as a copilot for freelance video editors, UGC creators, and independent content makers. By focusing on the "why" behind viewer drop-off, it seeks to provide a more sophisticated layer of feedback than standard platform insights provide after the fact.

How it works

The core technology involves a multi-step analysis of the uploaded video file. First, the AI transcribes the speech to understand the narrative flow and identifies the specific lines being spoken at any given time.

Next, the system measures technical elements like pauses in speech and the frequency of scene changes. It looks for patterns where the visual information might be static or where the spoken content becomes repetitive or overly long for the short-form format.

Finally, it compiles these observations into a list of notes. Each note is tied to a specific second in the video, providing the creator with a clear roadmap of what to trim, what to re-record, and how to structure the hook for maximum impact.