# Pollo Agent: turn a TikTok reference into a new, inspected video

Keep reference selection, creative decisions and generation quality visible throughout the workflow.

[HTML](https://lukejbyrne.com/library/guides/omcu3-2-_a0/)
Content language: en
Author: Luke Byrne
Video published: 2026-04-04
Guide published: 2026-11-02

[Watch the video](https://www.youtube.com/watch?v=Omcu3-2-_a0)
Video language: English

## Choose a reference and define what to borrow

The 4 April 2026 walkthrough uses a custom dashboard to find viral TikToks, then sends a reference to Pollo AI’s agent. The useful input is more than a popular link: identify the structure, visual idea or timing you want to explore. A reference’s popularity does not establish that a new version will perform similarly.

1. Use the retained scout repository: https://github.com/lukejbyrne/viral-tiktok-scout.
2. Follow the dashboard overview at 0:42 and select one reference for a specific creative reason.
3. Use the product route https://cutt.ly/ctP5qrFn and follow the reference-to-agent segment at 1:15.
4. Describe your new subject and what should change, rather than copying another creator’s identity or unsupported claims.

## Inspect the storyboard before accepting the clips

At 2:03 the recording iterates on storyboards and scenes; video generation and queued variants appear at 4:06. Keep those decisions separate. A storyboard can communicate a strong idea while a generated clip introduces inconsistent subjects, awkward movement or an ending that fails to deliver the setup.

1. Review the proposed scene order and check that each scene has a clear role in the new video.
2. Generate a small set of variants and inspect the beginning, transitions, subject consistency and ending of each.
3. Use the review at 5:10 and prompt refinement at 7:11 to target a visible weakness rather than request a vaguely better result.

## Automate only the production choices you understand

The API platform is introduced at 3:01. The description also names motion control, mini worlds, first-person POV and avatars, with additional features covered at 7:56. These are production options to choose for a purpose, not reasons to add every effect to one creative.

1. Retain the selected prompt, storyboard decisions and approved output together before repeating the process through the API.
2. Specify which inputs may vary and inspect the resulting batch against the original creative intent.
3. Keep generation automation separate from publishing or any later audience-performance test; this example reports a creation workflow, not measured viral success.

## The takeaway

The repeatable unit is a reference-informed brief with a reviewed output. Human selection and revision remain part of the method.

## Workflow resources

- [cutt.ly](https://cutt.ly/ctP5qrFn)
- [ailuke.short.gy](https://ailuke.short.gy/Omcu3-2-_a0_skool)
- [ailuke.short.gy](https://ailuke.short.gy/Omcu3-2-_a0_scripti)
- [github.com](https://github.com/lukejbyrne/viral-tiktok-scout)

## Chapters

- [0:00 — Finding viral TikToks to recreate](https://www.youtube.com/watch?v=Omcu3-2-_a0&t=0s)
- [0:42 — How the viral TikTok scout dashboard works](https://www.youtube.com/watch?v=Omcu3-2-_a0&t=42s)
- [1:15 — Sending a TikTok to the Pollo AI agent](https://www.youtube.com/watch?v=Omcu3-2-_a0&t=75s)
- [2:03 — Iterating on the storyboard and scenes](https://www.youtube.com/watch?v=Omcu3-2-_a0&t=123s)
- [3:01 — Using the Pollo API for automation](https://www.youtube.com/watch?v=Omcu3-2-_a0&t=181s)
- [4:06 — Generating videos and queuing multiple variants](https://www.youtube.com/watch?v=Omcu3-2-_a0&t=246s)
- [5:10 — Reviewing the finished AI videos](https://www.youtube.com/watch?v=Omcu3-2-_a0&t=310s)
- [6:01 — Why these viral formats actually work](https://www.youtube.com/watch?v=Omcu3-2-_a0&t=361s)
- [7:11 — Refining prompts for better results](https://www.youtube.com/watch?v=Omcu3-2-_a0&t=431s)
- [7:56 — Other Pollo features: motion control, mini worlds, avatars](https://www.youtube.com/watch?v=Omcu3-2-_a0&t=476s)

## Sources and limitations

Prepared 2 November 2026 from this exact published description and chapters. The recording was not independently watched or transcribed and no fresh hands-on test was performed. Steps are editorial adaptations; current setup, availability and offer terms may differ.

[Original video](https://www.youtube.com/watch?v=Omcu3-2-_a0)
