Seven Havens AI Prompt Guide
The scene does the directing — these notes make the take cleaner.
Start with the photo, not the prompt
Seven Havens AI is photo-first: the uploaded image anchors your face, so the shot you pick matters more than the words. A sharp portrait or half-body shot in even light gives the engine clean features to carry through the world.
- Face clearly visible: sunglasses, heavy shadows, and tiny faces confuse the likeness.
- Even light beats dramatic light — the engine re-lights for the scene anyway.
- Half-body shots give the bender wardrobe room to read; full-body works for skyline scenes.
- One or two people per photo — big groups dilute the elemental styling.
Let the scene template write the prompt
Every template already encodes the direction — the haven, the wardrobe, the camera move, the element. Open one and the prompt box fills with a tested brief; edit it only if you want a different moment.
- Haven Skyline: a rooftop over spirit-lit towers at dusk.
- Water Sanctuary: an arc of water rising to your hands under moonlight.
- Spirit Wilds: bioluminescent forest parting around you at night.
- Change one variable per run — the scene, the element, or the camera.
Writing a custom scene
When you do write your own, structure it like a shot list: person, place, action, camera, element. "Me on a haven rooftop at dusk, water bending arcing off my hands — fixed camera, cinematic anime light" gives the engine a complete brief.
- Name the move once: "water arcs off my hands" — not a whole fight scene.
- Anchor the camera: "fixed camera", "slow push-in", "drifting camera".
- Set the world cue: "haven world", "spirit-lit", "post-cataclysm" — the engine handles the rest.
- Preserve the face: "keep my face and expression unchanged".
Formats and sound
Pick the aspect for where the clip will run before you generate — 16:9 for YouTube, 9:16 for Reels and Shorts, 1:1 for the feed. Engines with the audio toggle can render ambience with the picture; otherwise add a score where you post.
- Vertical for TikTok/Reels/Shorts, landscape for YouTube and web embeds.
- Shorter clips post better — the bending reveal beats a long setup.
- Silent MP4 is normal — add music in your editor or the social app.
Reviewing the take
AI video varies between runs. Watch the face first — that is where drift shows — then the elemental FX and wardrobe edges. Keep the take where you still look like you, and label AI content where it matters.
- Check the face for softening or feature drift across the clip.
- Elemental FX glitches live at the hands — glance there second.
- Re-run the same scene when the take is close; change one thing when it is not.
Take the trip
One photo. One world: the havens.
Upload a photo and get a bender video of yourself you can post tonight.