WAN Bernini combined with Prompt Relay.
A workflow I use to create 10-15 second videos while being able to precisely time it through that length using Prompt Relay. With Bernini's innate ability to make longer, very consistent videos paired with prompt relay, precise control over segments as long as 15 seconds adhere to the prompt the closest to perfect I've ever seen.


This is the workflow's "soul".
Value stands for frame length (for some reason renaming it always defaulted back).
Type is used to switch between the default provided prompts for each of Bernini's 12 modes; r2v is reference image(s) to video, i2v is image to video, and so on.

Bernini is a unified framework for video generation and editing that combines an MLLM-based semantic planner with a DiT-based renderer.
More info:
https://huggingface.co/ByteDance/Bernini-R

Prompt Relay lets you split a text prompt into time-based segments so different actions happen at specific moments in a video without mixing up or breaking the visual style
More info:
GitHub - kijai/ComfyUI-PromptRelay · GitHub