Wan 2.7 Image & Video Models Now Available

We're excited to announce that both Wan 2.7 models are now available!

Wan 2.7-Image (April 1, 2026)

Lifelike Face Customization: Say goodbye to AI 'standard faces'. Wan 2.7-Image supports fine-grained facial customization including face shape (oval, round, square), eye features (almond, deep-set, phoenix eyes), and bone structure, creating truly unique 'thousand faces' portraits with real human warmth.

Color Palette Control: A new 'color palette' feature lets you extract color distributions from reference images with one click. Whether reproducing a master painting's palette or aligning with brand guidelines, the model delivers precise color migration while maintaining composition.

3K Token Ultra-Long Text Rendering: Supporting up to 3,000 tokens of text input across 12 languages, Wan 2.7-Image can render complex tables, mathematical formulas, and full A4 pages of academic content with print-level clarity.

Interactive Editing: Native interactive editing lets you simply select an area and add, align, or move elements with frame-level precision.

Multi-Subject Consistency: Supporting up to 9 reference images, the model maintains character and style consistency across serial content like storyboards, e-commerce photo sets, and multi-angle architectural views.

Batch Generation: Generate up to 12 images at once for creating consistent series, PPT illustrations, storyboard sequences, and multi-view architectural renders.

Wan 2.7-Image-pro, trained on larger-scale data, is also available with more stable composition and stronger semantic understanding.

Wan 2.7 Video Model (Late March 2026)

The Wan 2.7 Video model has launched on cloud platforms, delivering 1080p resolution, 15-second videos with native audio built in. Five key improvements over previous versions: ① Commercial-grade visual quality — sharper details, richer textures; ② Motion coherence — smoother, more realistic movement; ③ Native audio integration — sound is generated alongside video, not added after; ④ Style consistency — uniform aesthetics across the entire clip; ⑤ Temporal stability — no flickering or visual artifacts over time.

New video capabilities include: First/last frame control for precise start and end compositions; 9-grid image-to-video input for richer multi-angle scene composition; Subject + voice reference cloning for consistent character and voice across generations; Instruction-based editing to modify existing videos with text commands; Video recreation to re-generate videos in new styles while preserving structure.

Open-source release of the video model is expected mid-to-late Q2 2026. Credits on our platform never expire — unlike competitors whose credits reset monthly.