Build longer narratives with up to 30 seconds of output, more realistic detail, and cross-modal guidance from creative media or office materials.
Three generation modes cover direct prompting, controlled frame animation, and broad multimodal reference workflows.
Turn a prompt into a complete short sequence with selectable resolution, aspect ratio, sound, watermark, and seed controls.
Smart duration is intentionally unavailable during the beta so billing stays predictable.
Animate one image or guide the transition with an optional second image as the final frame.
Useful for product reveals, portrait motion, and directed transitions.
Combine images, short videos, audio, and one document or webpage to guide content and visual direction.
Each video or audio reference must be 1–15 seconds, with a 15-second total per media type.
Use presentations, spreadsheets, PDFs, text, Markdown, and Apple iWork files as creative context.
Document page-count validation is handled by the model provider.
A real beta image-to-video result generated through the same Wan 3.0 workflow available above.
Preserve the same elegant fashion subject and visual identity across the sequence. She steps into a luminous futuristic gallery as the camera makes a slow cinematic orbit; reflective surfaces, flowing fabric, precise facial detail, natural motion, refined commercial lighting, coherent colors, polished premium fashion-film finish.
Use the expanded duration for multi-beat product stories, cinematic establishing shots, educational sequences, and social videos that need more than a single moment.
Wan 3.0 aims for detailed motion and audiovisual output, while this beta clearly surfaces that voices, sound, visible text, and fine instructions may still vary between generations.
Wan 3.0 Video is a beta video generation model supporting text, first and last frames, and multimodal references including images, videos, audio, documents, and webpages.
Wan 3.0 costs 10 credits per second at 480p, 18 at 720p, and 36 at 1080p. Text and image modes bill output seconds; all-reference mode bills output plus reference-video seconds.
The integration is available for testing but may be less stable than production models. Sound quality, text rendering, instruction accuracy, and provider availability can still vary.
Supported formats are DOC, DOCX, XLS, XLSX, PPT, PPTX, PDF, TXT, Markdown, Keynote, Pages, and Numbers. Use one document or one webpage, not both.
Yes. Sound is enabled by default and may be disabled before generation. Audio quality and spoken or on-screen text accuracy may still fluctuate during the beta.
Compare beta multimodal creation, flagship production, LoRA-tuned character video, open-source workflows, multi-shot generation, and motion transfer without leaving the Wan family.
Start with text, a pair of frames, or a multimodal reference pack. Review the estimated credits before every submission.
