Precise generation control

WAN 2.5 Video Generator

Create AI videos with seed for reproducibility and negative prompts. Supports portrait and square formats.

5-10s
Duration
credits
~3-8 min
Generation
Seed
Reproducibility

WAN 2.5 unique features

Seed for reproducibility

Specify a seed for identical results. Perfect for iterations and A/B testing

DeterminismIterations

Negative prompt

Exclude unwanted elements: blur, artifacts, watermarks and other defects

Up to 500 charsFiltering

Three video formats

Create videos for any platform: YouTube, TikTok, Instagram and more

16:99:161:1

Generation modes

Text → Video

Create videos from text descriptions with synchronized audio and dialogue

5-10s720p/1080pNative audio

Image → Video

Bring a static image to life while preserving style, characters and composition

5-10sStable style

WAN 2.5 vs WAN 2.6

ParameterWAN 2.5WAN 2.6
Seed (reproducibility)
Negative prompt
Aspect ratios16:9, 9:16, 1:116:9 only
Auto-extend prompt
Video length5, 10s5, 10, 15s
Prompt limit800 charsNo limit
Video-to-Video
Multi-Shots mode
Price (min)

Prompt-writing tips

800-character limit

WAN 2.5 has a strict limit — maximum 800 characters per prompt. Our service automatically compresses prompts during translation, but writing concisely from the start is better.

  • Focus on the main action and a single scene
  • Use concrete nouns instead of descriptive adjectives
  • Enable Auto-extend — the API will fill in a short prompt

Optimal prompt structure

[Scene/Place] + [Subject] + [Action] + [Camera] + [Atmosphere]

"City at night, neon signs. A girl in a leather jacket walks on wet asphalt. Medium shot, slow camera pull-back. Rain, reflections in puddles."

Effective negative prompt

By default we use an optimized negative prompt, but you can change it:

blurry, low quality, distorted, watermark, text overlay, logo, grainy, pixelated, overexposed, underexposed, bad anatomy, deformed, ugly, artifacts, glitch, noise

Adding dialogue

WAN 2.5 supports lip-sync and audio synchronization. Write dialogue in quotation marks:

"A woman looks at the camera and says: «Привет! Я рада тебя видеть!» She smiles and waves. Warm light, blurred background."

💡 Dialogue in quotation marks is preserved as-is even after translation

Generation examples

Generation cost

Parameters720p1080p
5 seconds
10 seconds

💡 WAN 2.5 is cheaper than WAN 2.6 at the same parameters

Frequently asked questions

Q:How does WAN 2.5 differ from WAN 2.6?

WAN 2.5 supports seed for reproducible results, negative prompts to exclude unwanted elements, and three aspect ratios (16:9, 9:16, 1:1). However, WAN 2.5 has an 800-character prompt limit and does not support video-to-video mode or 15-second videos.

Q:What is Seed and why is it needed?

Seed is a number that determines generation randomness. With the same seed and the same prompt, you'll get an identical result. Useful for iteration: change the prompt while keeping the generation style.

Q:What is a negative prompt?

Негативный промпт описывает, чего НЕ должно быть в видео. Например: 'размытое, низкое качество, водяные знаки'. Это помогает избежать типичных артефактов AI-генерации.

Q:Why is the prompt limited to 800 characters?

Это техническое ограничение API WAN 2.5. Наш сервис автоматически сжимает промпты при переводе, сохраняя смысл. Также можно включить 'Авто-расширение' — тогда API сам дополнит короткий промпт.

Q:Which aspect ratios are supported?

WAN 2.5 supports three formats: 16:9 (landscape, for YouTube), 9:16 (portrait, for TikTok/Reels), and 1:1 (square, for Instagram). Unlike WAN 2.6, which only works with 16:9.

Q:Что делает функция 'Авто-расширение промпта'?

When this option is enabled, the API uses a built-in language model to automatically expand your short prompt. Especially useful given the 800-character limit.

Q:How long does generation take?

Generation typically takes 3 to 8 minutes depending on duration, resolution, and server load. A 5-second video at 720p generates fastest.

Q:Does WAN 2.5 support audio?

Yes! WAN 2.5 generates video with native synchronized audio. The model understands scene context and adds matching audio, including dialogue and ambient sounds.

Ready to try WAN 2.5?

Create AI videos with precise generation control. Seed for reproducibility, negative prompts, and three video formats.

Start generating

WAN 2.5 — AI model from Alibaba • Sixio

We use cookies to operate the service, keep your session, and collect anonymous statistics. See our Privacy Policy.