Wan 2.7 Edit Video: Natural-Language Video Transformation & Identity Preservation on playground and API | RunComfy

What makes Wan 2.7 Edit Video stand out

Wan 2.7 Edit Video is a video editing model built to modify existing footage with natural-language instructions while keeping identity, motion flow, and scene continuity stable. This edit-video task serves as a structure-aware mechanism for transforming source clips into revised outputs without discarding the original timing, subject presence, or visual logic. Wan 2.7 Edit Video is optimized for short-form footage, including talking-head content, product shots, and branded sequences where realism matters. Wan 2.7 Edit Video supports prompt-based edits with optional image reference control, and Wan 2.7 Edit Video is designed for concise, repeatable production workflows.

Key capabilities:

• Preserves subject identity and core scene structure during prompt-driven edits.
• Maintains motion consistency across short clips instead of treating frames independently.
• Supports reference-based editing through an optional reference image.
• Handles MP4 or MOV inputs from 2 to 10 seconds, up to 100 MB.
• Outputs in 720p or 1080p with selectable aspect ratios.
• Can match source duration or truncate from the start to a defined length.
• Offers audio control with either automatic handling or original audio preservation.

Prompting guide for Wan 2.7 Edit Video

Start Wan 2.7 Edit Video with a clean source clip and a direct instruction describing what should change and what must remain unchanged. For Wan 2.7 Edit Video, the most reliable prompts define the subject, the intended visual edit, and the constraints: identity, camera framing, motion pattern, and scene continuity. If using Wan 2.7 Edit Video with a reference image, specify whether the reference should guide style, appearance, or product detail. Keep prompts concrete and production-oriented. Wan 2.7 Edit Video works best when edits are scoped clearly, especially for short clips where temporal consistency is important. For predictable outputs in Wan 2.7 Edit Video, align resolution, aspect ratio, and duration with the source intent.

• "Preserve the speaker's face, pose, and lip movement; change the background to a modern office with neutral lighting."
• "Keep the product position and camera motion unchanged; restyle the scene into a luxury studio setup with soft reflections."
• "Maintain the original framing and hand movement; replace the packaging design using the reference image."
• "Preserve the subject and walking motion; convert the environment into a rainy nighttime street with realistic reflections."
• "Keep original composition and identity; apply a clean cinematic color grade and sharper commercial lighting."

Pro tips:

• State what to preserve first: subject identity, motion, framing, and timing.
• Limit each prompt to one primary edit direction plus a few hard constraints.
• Use reference images only when they directly support appearance or design transfer.
• Use a fixed seed when testing small prompt variations for comparison.
• Preserve source audio with "origin" when visual edits should not affect sound. For another edit-video workflow, use Wan 2.7 Edit Video playground.

Note: If you requires modifying an existing image - such as changing the background, lighting, or specific objects within a picture - please use the Seedream 4.5 Edit model, which is specifically optimized for instruction-based image manipulation.