Alibaba Unveils Wan2.7-Video to Elevate Creators from Executors to Directors

By evolving AI from simple asset generation to full finished-film creation, Wan2.7-Video provides director-level control over entire workflows, allowing creators to orchestrate complex narratives and post-production with unprecedented precision.

This video model release comes just days after the debut of its latest image generation model Wan2.7-Image, marking a rapid, coordinated expansion of Alibaba's multimedia AI capabilities.

Mastering the Full Creative WorkflowThe Wan2.7-Video features four models including Text-to-Video model Wan2.7-t2v, Image-to-Video Wan2.7-i2v, Reference-to-Video Wan2.7-r2v, and Video Editing Wan2.7-videoedit. It unifies text, image, video, and audio inputs into a single cohesive system covering generation, editing, replication, reshaping, continuation, and referencing.

This comprehensive suite addresses critical gaps in narrative coherence and multi-shot consistency, enabling users to handle diverse scenarios from script input to image control effortlessly.

Wan2.7-Video supports flexible video generation ranging from 2 to 15 seconds and delivers 720p and 1080p output. Its enterprise APIs enable batch processing and tailored workflows.

Director-Grade Editing and ControlAddressing the high cost of traditional post-production, Wan2.7-Video series introduces intuitive natural language instructions that allows creators to modify any video dimension, including character actions, dialogue, appearance, scenes, styles, and shooting methods. Users can redefine cinematography quickly with dozens of basic and complex camera movements while maintaining perfect lighting consistency.

The model dynamically edits dialogues by auto-syncing lip movements and preserving unique vocal signatures, even when scripts are rewritten. Multimodal inputs enable granular control, such as using audio clips to transform weather conditions or multi-panel images to precisely dictate story direction, camera composition, and character settings.

The system ensures cross-video consistency for up to five distinct characters by customizing exclusive voice tones and visual identities for each subject across complex narratives. Additionally, the model supports thousands of style combinations and over 50 refined emotional expressions for realistic performance.

Intelligent Cinematic StorytellingThe model's narrative engine transforms simple inputs into dramatic, interlocking storylines through intelligent scheduling. A single prompt can yield a fully realized storyboard featuring professionally construction and multi-storyboard paced transitions, FPV (first-person view) drone dives, 360-degree orbital shots, and context-aware lighting.

Trained on aligned video edits, Wan2.7-Video utilizes a novel end-to-end architecture for precise content reconstruction via natural language and reference images. Its video continuation mechanism allows users to specify ending frames, ensuring smooth transitions and eliminating abrupt "braking" effects.

Wan2.7-Image to Bring Unprecedented Personalization and Color AccuracyShortly before the release of its Wan2.7-Video model, Alibaba introduced the Wan2.7-Image model, a transformative visual generation tool designed to overcome the generic aesthetics and inconsistent color reproduction that have long plagued AI imagery. By delivering unprecedented personalization and professional-grade precision, the model outperforms leading industry players in blind human tests for visual fidelity, text rendering, and concept understanding.

Key innovations include a deep personalization engine that allows users to fine-tune specific character traits like bone structure and eye shape, alongside a unique "color palette" feature that enables exact color code matching for brand consistency. The model also achieves a major breakthrough in text rendering by leveraging a 3,000-token context window to generate print-quality academic text, complex formulas, and tables across 12 languages, effectively bridging the gap between linguistic intent and visual execution.

To maximize creator efficiency, Wan2.7-Image supports high-volume workflows by processing up to nine reference images and generating 12 distinct outputs in a single batch, streamlining the production of storyboards and e-commerce campaigns. This generative power is complemented by an intuitive "click-to-edit" interface that offers pixel-level control for adding, moving, or aligning elements with remarkable accuracy. Alibaba simultaneously launched Wan2.7-Image-Pro, an enhanced version featuring sharper prompt interpretation, more stable composition, and high-definition 4K output.

Both the video and image models are now available on Alibaba Cloud's Model Studio and the official Wan website, with integration into the Qwen App, marking a significant evolution in the Wan series' commitment to providing creators with intuitive, high-performance multimedia solutions.

• ข่าวo:member
• ข่าวo:loc
• ข่าวen
• ข่าวo:it