Explore the latest video effects
BlogVirvy
—
AI Video2026-10-20•7 min read

Wan 2.7 Video Prompts: Settings for Clearer First Tests

A practical Wan 2.7 guide with supported duration, resolution, aspect-ratio, audio, and reference-image choices, plus original prompts to copy and adapt.

Editorial illustration; this is not a Wan 2.7 output.

Start with one action and one camera move

Wan 2.7 can be used for text-to-video and reference-driven video workflows. For a first test, describe one visible action, one setting, and one camera movement. This makes it easier to see which part of the prompt needs changing when the result misses the brief.

The prompts below are original suggestions. They have not been run or benchmarked as verified outputs, so treat them as starting points rather than guarantees about motion, identity, audio, or text.

  • Text or reference operation: choose a duration from 2 to 15 seconds; the default is 5 seconds.
  • Resolution: choose 720p or 1080p. The text/reference default is 720p; image-first-frame generation defaults to 1080p.
  • Aspect ratio: text/reference supports 16:9, 9:16, 1:1, 4:3, and 3:4; the default is 16:9.
  • The current API contract enables generated audio by default. Final availability and the quote depend on the product surface you use; this guide does not promise an audio toggle.

Choose the right reference mode

If you provide a first frame, the image operation requires one first-frame image and accepts an optional last frame and reference audio. Its aspect ratio is adaptive, and a 2–15 second duration is supported. Use this mode when the opening image matters more than a fixed canvas ratio.

The underlying reference operation accepts a total of one to five references, with up to five images and up to three videos; a reference video can be up to 10 seconds. The current Virvy composer flow accepts one uploaded reference photo, so follow the controls shown in the product. Only upload media you have permission to use, and do not assume a reference will preserve every face, logo, or detail exactly.

A simple iteration loop

Render a short draft first. If the subject moves correctly but the composition is wrong, change the camera sentence before changing the whole prompt. If the opening is right but the ending drifts, try a first-frame or first-and-last-frame workflow. Keep the original input and settings beside each result so you can compare changes.

Copyable prompt ideas

Text to video · slow product reveal

A small red ceramic cup sits on a pale stone table beside a window. Morning light moves across the table as the camera makes one slow push-in from a medium shot to a close-up. Steam rises gently from the cup. Keep the table, cup shape, and window position consistent throughout the shot.

Suggested settings: 5 seconds · 16:9 · 720p · generated audio enabled

Text to video · vertical street scene

A cyclist in a yellow rain jacket rides slowly through a quiet city street after rain. The camera tracks from the left at walking speed while reflections pass across the wet pavement. Keep the rider centered and show one clean turn at the end. Natural street ambience, no dialogue.

Suggested settings: 6 seconds · 9:16 · 720p · generated audio enabled

Reference image · gentle portrait motion

Use the uploaded portrait as the first frame. The person turns their head slightly toward the window and smiles once while soft curtain light changes across the background. Keep the framing, clothing, and facial features close to the uploaded image. Use a subtle handheld camera move.

Suggested settings: First-frame image · 4 seconds · adaptive ratio · 1080p

Reference image · paper airplane motion

Use the uploaded desk photo as the first frame. A paper airplane lifts from the desk and travels toward the open window in one smooth motion. Move the camera gently with the airplane and keep the room layout, desk, and window recognizable throughout the shot.

Suggested settings: One first-frame image · 5 seconds · adaptive ratio · 1080p

Sources and verification

Model settings were checked against these sources on 2026-10-20. Prompt examples on this page are original suggestions and were not presented as verified model outputs.