How to make longer AI videos with consistent image/model?

Veo3 is great for top quality audio+video but the output is limited to 8 seconds. The "extend" feature in Google Flow doesn't work with Veo3 yet. So if you want to go beyond 8 seconds you need to generate a new video and pray it looks the same.

Has anyone figured a work around to break past the 8 second mark? Are there other platforms that could help with this? Haven't tried Higgsfield yet

Every time you make a prompt, first phrase of your prompt should be the same description of the characters or/and the background.
"Black policeman with black curly hair wearing a sunglasses on the beach..."
Maybe there are more options.
Note if you copy past the same prompt it should be the same or something close, right? Then you can add slight difference at the end of the structure

First, create an 8 second Veo3 clip and save its final frame as an image. Next start another 8 second clip,, using that saved frame as the first frame while keeping the same seed and prompt, then add one new action. Last, stitch the clips together in a video editor, the join is so smooth you can hardly notice it.

The extend part is not supported yet, so it is quite difficult. The way I do it is to create each 8 second segment with the same prompt and seed to keep the image consistent. Then combine the videos using editing software. This method is a bit manual, but it helps avoid characters or scenes being too misaligned. In addition, preparing the prompt carefully, keeping the lighting, style, and background parameters stable also helps the video be more seamless when combining the segment