Skip to content
IFAI

Generative media

Alibaba's Wan3.0 turns PPTs into 30-second AI videos

Published Reach 42

Alibaba Cloud just launched Wan3.0, extending single-pass video generation to 30 seconds. For the first time, the model can ingest office files like PDFs, slides, and spreadsheets alongside text and images to produce narrative videos. It also sharpens facial fidelity and adds built-in video editing, with API pricing starting at $0.05 per second.

AI video generators have spent years giving creators high-res fragments instead of real stories. Sticking together four-second clips to tell a coherent narrative usually feels like wrestling with a slot machine that keeps forgetting what your main character looks like.

Alibaba Cloud just launched Wan3.0 on its Model Studio and Qwen platforms. The big leap here is twofold: single-generation video duration now stretches to a full 30 seconds, and you can literally drop entire documents like PDFs, PPTs, and spreadsheets directly into the prompt box.

Documents in, commercials out

Most video models accept text and occasionally a reference image. Wan3.0 opens the gates to docs, PDFs, spreadsheets, Markdown files, and web links, supporting files up to 50 pages and 100MB. Pair a brand deck with a short sentence describing the mood, and the model orchestrates the content into a cohesive commercial.

Duration-wise, Wan3.0 doubles the 15-second cap of its predecessor, Wan2.7. It introduces smart duration recommendations to pick the right pacing based on your prompt, along with timeline extensions for long-form narrative. It also keeps the iterative video editing features introduced in Wan2.7, letting you tweak specific visuals, storylines, or dialogue without re-rolling the whole generation from scratch.

Alibaba also focused heavily on facial realism, avoiding the generic, over-smoothed aesthetic typical of AI avatars in favor of restrained micro-expressions and consistent spatial depth. On the flip side, Alibaba openly notes that audio textures and crisp text rendering within the video frames are still work in progress.

API pricing is billed by the second: $0.05 per second for 480P, $0.10 for 720P, and $0.20 for 1080P. A 30-second 1080P render comes out to $6.00, while a rough 480P draft costs $1.50.

Why it matters

For solo creators, agencies, and marketing teams, 30 unbroken seconds turns AI video from a B-roll generator into a genuine production tool. Continuous camera tracking, one-take shots, and narrative arc can finally exist within a single generation call. Document-to-video also bypasses the tedious process of turning pitch decks into storyboard prompts by hand.

From an industry perspective, Alibaba is attacking the video race from a distinctly enterprise angle. While rivals compete on photorealistic 5-second snippets, Alibaba is connecting multi-page business data directly to video output, positioning Wan3.0 as a universal information layer rather than just an artistic engine.

We think the tiered pricing structure is particularly smart. Charging $1.50 for a low-res draft lets teams iterate on pacing and camera moves affordably before committing $6.00 to high-res rendering. That said, sustaining coherence over 30 full seconds is notoriously tough; if spatial logic breaks halfway through, creators still face editing headaches. Combined with the acknowledged weaknesses in on-screen text and audio quality, Wan3.0 won't eliminate post-production suites just yet, but it dramatically shortens the path from rough concept to finished cut.

Sources (2)

  1. [1]
    alibabacloud.com(opens in a new tab)Originalalibabacloud.com
  2. [2]
    alibabacloud.com(opens in a new tab)Reportalibabacloud.com

Best of AI, daily