Wan 2.7 — A Full Upgrade for AI Video Generation
Wan 2.7 is the latest iteration of the Wan video generation model, delivering meaningful improvements in character consistency, physical motion, and prompt comprehension. It supports 5–15 second video generation at both 720p and 1080p resolution, making it a practical choice for everything from quick social clips to polished marketing content.
What's New in Wan 2.7
Compared with Wan 2.6, the upgrades are not incremental — they address the core pain points that creators encounter most frequently.
Character Consistency
Wan 2.7 maintains stable facial features, body proportions, and clothing details across every frame. Characters no longer "drift" mid-clip — even in complex multi-action sequences.
Smoother Physical Motion
Motion dynamics are more physically plausible. Acceleration, deceleration, gravity, and object interactions feel natural rather than artificial, especially in walking, running, and hand gesture sequences.
Better Prompt Understanding
The model parses longer, more detailed prompts with higher fidelity. Spatial relationships, sequential actions, and nuanced scene descriptions are interpreted more accurately than previous versions.
720p & 1080p · 5–15 Seconds
Generate videos from 5 to 15 seconds in length at 720p or 1080p resolution. Shorter clips maximize per-frame quality; longer clips suit narrative and cinematic use cases.
How to Write Prompts for Wan 2.7
A well-structured prompt is the single most important factor in getting high-quality, consistent output from Wan 2.7. We recommend organizing your prompt into three distinct parts: Subject, Description, and Camera Language.
Part 1 — Subject
The subject line tells the model who or what is in the video. When working with a reference image — especially a character — it is critical to include identity-locking phrases to prevent the model from altering facial features or body proportions.
Example
the same beautiful young woman, identical face and body proportions as referenceWhy this matters
Adding the same ... or identical face explicitly anchors the model to the reference image's identity. Without these phrases, the model may generate a similar-looking but noticeably different person — subtle changes in jawline, eye shape, or proportions that break continuity across clips.
Part 2 — Description
This is the most important part of your prompt. The description tells the model what is happening — the scene, actions, poses, expressions, environment, and any changes over time. Be specific and sequential.
Scene & Environment
“standing on a rooftop terrace at golden hour, city skyline in the background, warm orange light”
Action & Movement
“slowly turns to face the camera, wind gently blowing through her hair”
Pose & Gesture
“one hand resting on the railing, the other brushing hair behind her ear”
Expression & Emotion
“soft confident smile, relaxed eyes, calm and composed demeanor”
The more spatial and temporal detail you provide, the more faithfully the model reproduces your creative intent. Vague prompts like “a woman standing outside” leave too much room for interpretation.
Part 3 — Camera Language
Camera language controls the cinematographic framing and feel of the output. Include shot type, angle, and depth of field to guide composition.
Example
medium close-up, low angle, shallow depth of field| Category | Options |
|---|---|
| Shot Type | extreme close-up · close-up · medium close-up · medium shot · full shot · wide shot |
| Angle | eye level · low angle · high angle · dutch angle · bird's eye · worm's eye |
| Depth of Field | shallow depth of field (subject in focus, blurred background) · deep focus (everything sharp) |
Full Prompt Example
Here is a complete prompt that combines all three parts — subject, description, and camera language — into a single, production-ready input.
The same beautiful young woman, identical face and body proportions as reference, wearing a cream-colored knit sweater and light blue jeans.
She stands on a sunlit wooden pier by the ocean at golden hour. She slowly walks toward the camera with a relaxed, confident stride, the ocean breeze lifting strands of her hair. She pauses, tilts her head slightly, and breaks into a warm, genuine smile. Waves crash softly in the background, and the warm light casts a soft glow across her face.
Medium close-up, low angle, shallow depth of field, golden backlight with soft lens flare.
Subject anchoring
"the same ... identical face" locks identity to the reference image
Rich description
Sequential actions (walks → pauses → tilts → smiles) give the model temporal structure
Cinematic camera
Low angle + shallow DoF + backlight creates a professional, editorial look
Tips for Best Results
Always Include Identity-Locking Phrases
When using a reference image, always add "the same ..." or "identical face and body proportions as reference" in your subject line. This is the single most effective way to prevent character drift and maintain consistency.
Describe Actions in Sequence
Instead of describing a static scene, write actions in the order they should happen: "she walks forward, pauses, and turns to face the camera." This gives the model a clear temporal roadmap for the 5–15 second clip.
Be Specific About Environment
Include lighting conditions (golden hour, overcast, neon-lit), setting details (cobblestone street, glass office, sandy beach), and atmospheric elements (fog, rain, wind). These details dramatically affect output quality.
Match Duration to Complexity
For a single action with minimal scene change, 5 seconds is usually sufficient. For sequences with multiple actions or camera movements, use 10–15 seconds to give the model enough temporal space.
Choose Resolution by Use Case
720p renders faster and works well for drafts, social stories, and rapid iteration. 1080p is best for final output — marketing videos, portfolio pieces, and any content that will be viewed on larger screens.
Frequently Asked Questions
What video lengths does Wan 2.7 support?
Wan 2.7 generates videos between 5 and 15 seconds. Shorter clips (5s) are great for quick social content, while longer clips (10–15s) are better suited for narrative sequences and cinematic storytelling.
What resolutions are available?
Wan 2.7 supports both 720p and 1080p. Use 720p for faster iteration and drafts. Use 1080p for final, production-quality output.
How do I keep the character looking the same across clips?
Include identity-locking phrases in your prompt's subject line, such as "the same beautiful young woman, identical face and body proportions as reference." This anchors the model to the reference image and prevents facial or body drift.
What makes Wan 2.7 better than Wan 2.6?
Wan 2.7 delivers significantly improved character consistency, smoother and more physically plausible motion, and better understanding of complex, detailed prompts. The overall output quality is a noticeable step up from 2.6.
How detailed should my prompts be?
The more detail, the better. We recommend structuring prompts into three parts: Subject (who/what), Description (scene, action, pose, expression), and Camera Language (shot type, angle, depth of field). Vague prompts produce generic results.
Can I use reference images with Wan 2.7?
Yes. Wan 2.7 supports reference image input. When using a reference, always pair it with identity-locking language in your prompt to ensure the generated character matches the reference as closely as possible.