Wan 2.6 Text-to-Video Spicy API

Wan 2.6 Text-to-Video Spicy is Alibaba Cloud's next-generation video generation model that produces high-definition video from text prompts. The Wan 2.6 series was officially released on December 16, 2025, extending maximum video duration to 15 seconds with native audio-visual joint generation (sound effects, background audio, and multi-language lip-sync). Per the official specification, a single text-to-video generation runs up to 15 seconds, delivering comprehensive upgrades over the previous generation in visual quality, instruction adherence, intelligent multi-shot storytelling, and audio-visual synchronization. This model supports 720P and 1080P resolution tiers (five aspect ratios each), 5/10/15-second duration options, and an optional watermark.

The model is available via the iCreat platform API using a two-step asynchronous task interface: submit a task to obtain a task_id, then poll the result endpoint until a terminal state is reached. Billing is based on output video duration: 720P at $0.10/second and 1080P at $0.15/second.

Model Positioning

Wan 2.6 Text-to-Video is positioned as the flagship long-duration text-to-video tier in the Tongyi Wanxiang family. It handles the core task of generating video directly from text: compared to Wan 2.2's fixed 5-second lightweight tier, it extends duration to 5/10/15 seconds and pushes resolution up to 1080P, covering a broader range from quick clips to full narratives. It complements the series' Reference-to-Video (R2V) model — text-to-video creates from scratch, while R2V brings existing characters into new scenes. The 10- and 15-second tiers let a single video carry a complete shot sequence, and combined with the official intelligent multi-shot storytelling capability (generating coherent multi-shot narratives from a single prompt), it is well-suited for commercial ads, short dramas, and social media content that require narrative completeness.

Core Capabilities

Three Duration Tiers for Flexible Storytelling

Duration supports 5/10/15 seconds: 5 seconds for fast-paced clips, 10 seconds for a complete performance, 15 seconds to tell a short story — all covered through one interface from fragments to narratives.

Dual Resolution Tiers with Ten Aspect Ratios

720P and 1080P resolution tiers, each with five aspect ratios (16:9, 9:16, 1:1, 4:3, 3:4) — landscape, portrait, and square all covered, delivery-ready dimensions in one step.

Multi-Shot Storytelling

The series supports intelligent multi-shot storytelling: converting simple prompts into multi-shot scripts to generate coherent narrative videos with multiple shots, maintaining consistency of key elements such as subjects and scenes across shots.

Native Audio-Visual Synchronization

Per the official specification, the series natively integrates audio-visual joint generation: sound effects, background audio, and multi-language lip-sync are produced by the model, with audio-visual alignment built into the model rather than post-production stitching.

Improved Instruction Adherence

Compared to the previous generation, instruction adherence precision and visual quality are improved, complex prompts are executed more reliably, and the probability of visual artifacts and deviations is reduced.

Pricing

Total cost = unit price × output video duration. The size value is billed according to its resolution tier: 1280*720, 720*1280, 960*960, 1088*832, 832*1088 belong to the 720P tier; 1920*1080, 1080*1920, 1440*1440, 1632*1248, 1248*1632 belong to the 1080P tier. The costUSD field in the response returns the actual cost upon task success.

Application Scenarios

• Commercial advertising: 10–15-second full-narrative brand films and product showcases
• Short dramas and storytelling: multi-shot coherent narrative shot content
• Social media video: vertical short videos and topical content batch production
• Creative exploration: iterating across duration and aspect ratio combinations
• Dynamic asset production: automated content pipeline from copy to finished footage

Model Comparison

Same-Series Model Comparison

Cross-Model Comparison

Same-series data is from the official Tongyi Wanxiang release; this model's specifications are per the iCreat channel documentation.

Why Choose Wan 2.6 Text-to-Video?

• 5/10/15-second three-tier duration, from fast-paced clips to full narratives in one interface
• 720P / 1080P dual-tier ten aspect ratios, delivery-ready dimensions in one step
• Multi-shot intelligent storytelling, single video carrying complete shot language
• Native audio-visual joint generation, audio-visual alignment built into the model
• 720P at $0.10/second, a 10-second finished video for $1.00 — controllable cost for narrative-grade content

Specifications

Architecture

Wan 2.6 is Alibaba Cloud's Tongyi Wanxiang video generation model series released on December 16, 2025, covering text-to-video, image-to-video, reference-to-video, image generation, and text-to-image. The series extends text-to-video duration to a maximum of 15 seconds, with native audio-visual joint generation — sound effects, background audio, and multi-language lip-sync produced by the model; it also introduces intelligent multi-shot storytelling, converting simple prompts into coherent multi-shot narratives while maintaining cross-shot consistency. Compared to the previous generation, instruction adherence precision and visual quality are both improved, with text and fine detail retention being more stable. Text-to-Video is the series' flagship T2V endpoint, and this channel provides 720P and 1080P resolution tiers with 5/10/15-second duration options.

Notes

• Submit and query must use the same task_id to chain the operation
• While processing, result is [] — continue polling, recommended interval 2–5 seconds
• FAILED is a terminal state — check request parameters and resubmit
• duration only accepts 5, 10, or 15 — other second values are not supported
• size uses width*height format (e.g., 1280*720), not x as separator
• size must be one of the ten valid values — custom dimensions will be rejected
• input and parameters are top-level sibling fields — do not nest them
• costUSD is returned only on task success; failed tasks do not produce a cost field

FAQ

How do I get started with the Wan 2.6 Text-to-Video API?​

Register on the iCreat platform and obtain an API Key from the console. Send a generation request to the submit endpoint, then use the returned task_id to poll the result endpoint. All requests are authenticated via the Authorization: Bearer header — keep your API Key secure and avoid exposing it in client-side code or public repositories.

How is billing calculated?​

Total cost = unit price × output video duration, billed by the resolution tier of the size value. 720P is $0.10/second: 5 seconds = $0.50, 10 seconds = $1.00, 15 seconds = $1.50; 1080P is $0.15/second: 5 seconds = $0.75, 10 seconds = $1.50, 15 seconds = $2.25. The costUSD field in the response provides the actual cost upon task success.

How do I choose a size?​

The ten valid values are split into two resolution tiers with five aspect ratios each: 720P tier 1280*720 (16:9), 720*1280 (9:16), 960*960 (1:1), 1088*832 (4:3), 832*1088 (3:4); 1080P tier 1920*1080, 1080*1920, 1440*1440, 1632*1248, 1248*1632. Choose the aspect ratio by your delivery platform and the tier by your clarity requirements.

Can I set an arbitrary duration?​

No. duration only supports three enumerated values: 5, 10, and 15 (default 5). Intermediate durations such as 7 or 12 seconds are not supported. For scenarios requiring continuous 4–15-second durations, consider Wan 2.6 Reference-to-Video or Seedance 2.5 Economy.

How do I choose between this and Wan 2.2 Text-to-Video?​

Two paths: if you need full narratives (10/15 seconds), 1080P high definition, and multi-shot storytelling — choose this model; if you only need 5-second fast-paced drafts at the lowest cost — choose Wan 2.2 Text-to-Video (480P tier at $0.02/second, roughly one-fifth of this model's 720P rate).

What should I do if a task fails?​

FAILED is a terminal task state. Check in order: is prompt empty, is size one of the ten valid values in width*height format, is duration one of 5/10/15, and are input and parameters top-level sibling fields? Fix the issue and resubmit the task. The cost field is returned only on task success.