How to transform a video clip with AI
Beyond generating video from scratch, Gemini Omni can take an existing video clip and transform it — restyle it, change the scene, or push it in a new creative direction, all from a text prompt. This "video-to-video" mode is one of Omni's most useful features. Here's how to use it on Dora.
What video-to-video is good for
- Restyling footage — turn a plain clip into a cinematic, animated, or stylized version.
- Scene transformation — change the setting or mood while keeping the action.
- Creative remixes — reinterpret a moment in a completely different aesthetic.
Step 1 — Select Gemini Omni and add a clip
Open the create page, choose Gemini Omni, and upload a video. Dora uses the first 10 seconds of your clip as the transformation window, so trim to the moment you care about (files up to 100 MB).
When a video is present, Dora automatically switches to video-input mode: the duration selector disappears (output length follows the model) and pricing becomes a flat per-generation rate by resolution.
Step 2 — Describe the transformation
Your prompt is the instruction for how to change the clip:
"Transform into a hand-painted watercolor animation, warm palette, soft edges, keep the same motion."
Be explicit about what to keep ("same motion", "same subject") and what to change ("new background", "anime style").
Step 3 — Choose resolution
Pick 720p, 1080p, or 4K. Resolution sets both quality and price in video mode. Start at 720p to dial in the prompt, then re-run at higher resolution for the final.
Step 4 — Generate, compare, refine
Run it, compare against your source, and adjust. If the transformation is too strong, soften the prompt; if it's too subtle, lean harder into the style description.
Tips
- Short, clean clips work best — busy, fast-cut footage is harder to transform coherently.
- One idea per prompt. "Watercolor + neon + slow-mo + new city" fights itself; pick the dominant look.
- Stabilize first. Shaky source video carries through.
Other ways to animate
Want to apply a specific motion (like a dance) to a still photo instead? That's Dora's Motion Control model. To learn the broader picture, read what Gemini Omni is.