How to Edit Existing Video With AI with Runway Aleph

Aleph is V2V-only and requires a video input. Route the original campaign clip into a video reference node on Astorie's canvas. The model will preserve the source's camera move, character action, and timing — what changes is the look layer.

Aleph's look transfer is driven primarily by an image reference, not by text. Drop a separate image reference (a still that captures the new season's palette, lighting, mood). For winter restyle: a snowy moodboard photo. For dusk pivot: a blue-hour image. The cleaner and more aesthetically defined the reference, the cleaner the restyle.

Aleph's prompt is for tone direction, not scene composition. "Restyle to winter early morning, soft snow underfoot, breath visible, palette shifts to cool blues and whites, character wardrobe layers heavier coats." Keep prompts short and shift-oriented. Aleph reads images more than language; the prompt is the steering wheel, the image reference is the engine.

For a seasonal pivot where the brand spokesperson must remain identifiable, explicitly tell Aleph what to keep: "preserve character face and brand-color logo, restyle only environment and lighting." Without this guard, Aleph will sometimes restyle wardrobe to match the season too aggressively. The directional prompt is your contract.

Aleph renders take 2-4 minutes per pass. On the canvas, render 2-3 variants with different look references (e.g. three winter moodboards: pine-forest, urban-snow, alpine-cabin). Place them side-by-side with the source. Pick the winner before going to NLE export. Cheaper than rendering on a single guess.

Once the winning Aleph restyle is locked, route it into the sequence builder, layer audio (often the original audio works because timing is preserved), and export as a native sequence to Premiere, DaVinci, or Final Cut. The seasonal campaign deliverable matches the original cut precisely on timing, just in the new look.