A new text-and-image-to-video model
A short rulebook for getting good clips out of Boreal.
Try BorealHow to best prompt with Boreal.
Phrases that reliably move the output.
Add forehead relaxed and smooth to avoid wrinkle artifacts on close talking shots.
Say the voice is close, warm, and full, never thin or tinny. Boreal generates audio jointly, so describe it.
For products, add keeps its exact shape, color, and text. Holding the product intact is one of Boreal's biggest strengths.
Natural room tone, no music keeps UGC clips clean; otherwise describe the ambience.
End with no captions, no logos, no added text.
Five clips the VDE team produced on Boreal, each with the exact prompt and source image behind it. Copy, paste, swap in your own details.
What to add when a take comes back wrong.
| Symptom | Add to the prompt |
|---|---|
| Motion too fast | unhurried, relaxed pace |
| Face morphs or brow wrinkles | forehead calm and still |
| Tinny or thin audio | voice close, warm, and full, not tinny |
| Product warps or text scrambles | holds its exact shape, color, and text |
| Camera drifts | the camera stays completely static and locked, no zoom, no pan |
| A take just misses | regenerate and make several, then pick a winner |
Set the scene with your image, describe only what moves in one flowing paragraph, lock the camera, keep it around 8 seconds, and generate ten to keep the best.
Build with Boreal