Set Duration and Resolution Every Time
Both are required on both models: duration takes the string "5" or "10", and resolution takes 720p or 1080p.
Input
The text prompt describing the desired video motion
0/800Click to upload or drag and drop
Supported formats: JPEG, PNG, WEBP Maximum file size: 10MB
URL of the image to use as the first frame. Must be publicly accessible
The duration of the generated video in seconds
Video resolution. Valid values: 720p, 1080p
Negative prompt to describe content to avoid
0/500Whether to enable prompt rewriting using LLM
Random seed for reproducibility. If None, a random seed is chosen
A configurable parameter. Defaults to true in the Playground.
There is no guarantee that everything can be filtered out; if you are not satisfied with the results, you will need to make your own arrangements.
Output
output typevideo
Examples
Explore different use cases and parameter configurations
Input
The text prompt for video generation. Supports Chinese and English, max 800 characters.
0/800The duration of the generated video in seconds
The aspect ratio of the generated video
Video resolution tier
Negative prompt to describe content to avoid. Max 500 characters.
0/500Whether to enable prompt rewriting using LLM. Improves results for short prompts but increases processing time.
Random seed for reproducibility. If None, a random seed is chosen.
A configurable parameter. Defaults to true in the Playground.
There is no guarantee that everything can be filtered out; if you are not satisfied with the results, you will need to make your own arrangements.
Output
output typevideo
Examples
Explore different use cases and parameter configurations
README
Complete guide to using Wan 2.5
Wan 2.5 API: Text-to-Video and Image-to-Video With Audio
Wan 2.5 Preview is Alibaba's video model that generates audio together with the video, in clips up to 10 seconds long. On Kie.ai it runs as two API models: wan/2-5-text-to-video and wan/2-5-image-to-video.
Wan 2.5 API Parameters on Kie.ai
Both models use POST /api/v1/jobs/createTask, and results come back by callBackUrl or Get Task Details.
| Parameter | Text-to-video | Image-to-video |
|---|---|---|
| Model | wan/2-5-text-to-video | wan/2-5-image-to-video |
| Prompt | Required; up to 800 characters; Chinese or English | Required; up to 800 characters |
| Start image (image_url) | Not used | Required; a publicly accessible JPEG, PNG or WebP URL, up to 10 MB, used as the first frame |
| Duration | Required; "5" or "10" seconds | Required; "5" or "10" seconds |
| Resolution | Required; 720p or 1080p | Required; 720p or 1080p |
| Aspect ratio | 16:9, 9:16 or 1:1 | Follows the start image |
| Output size in our tests on Kie.ai | 1920×1080 at 1080p, 16:9; 720×1280 at 720p, 9:16 | 1440×1440 at 1080p from a square image; 1110×830 at 720p from a 4:3 image |
| Audio track in our tests on Kie.ai | Yes, in both tests | Yes, in both tests |
| Optional | negative_prompt (up to 500 characters), enable_prompt_expansion, seed | negative_prompt (up to 500 characters), enable_prompt_expansion, seed |
| Pricing | Shown in the pricing card above | Shown in the pricing card above |
What Wan 2.5 Can Generate
Each image below shows three frames (start, middle and end) from one video we generated with Wan 2.5 on Kie.ai in October 2026 from the prompt described in its caption; the videos also have sound, which still frames cannot show.
Text-to-Video With Sound: Rain on a Cafe Window at Night
A 10-second 1080p landscape video: rain runs down the glass, the city lights outside stay blurred, and the camera pushes in toward the window until the cup on the table moves out of frame. The prompt also described the sounds (rain on glass, distant traffic), and the video came back with an audio track.
Vertical Video: Lifting Ramen Noodles
A 5-second 720p vertical video: chopsticks lift noodles out of the bowl while broth drips and the camera tilts up. The bottles and packets in the background carry made-up lettering, and a second bowl appears in the foreground partway through.
Image-to-Video: Our Robot Mascot Waves
Starting from our robot image as the first frame, Wan 2.5 made a 10-second 1080p video of the robot waving at the camera. Its antenna lights changed from yellow to pink along the way, which the prompt did not ask for.
Image-to-Video With Text in the Picture: A Chalkboard Menu
From a chalkboard menu image, the prompt asked for the chalk steam to rise while the lettering stayed still. By the 2.5-second mark the prices had disappeared and the cup drawing had changed shape, while the dish names stayed.
Which Wan Model to Use
Alibaba describes Wan 2.5 Preview as supporting synchronized audio generation with visuals and 10-second videos. Wan 2.2, the earlier generation, has open weights that Alibaba has published on Hugging Face, so you can run it yourself. A practical default is Wan 2.5 when you want sound generated with the video through an API, and the earlier open model when you need weights to host yourself.
| Wan 2.5 Preview | Wan 2.2 (A14B, open weights) | |
|---|---|---|
| Audio generated with the video | Yes (Alibaba; also in our tests on Kie.ai) | Not described |
| Open weights | Not on Wan-AI's Hugging Face page (October 2026) | Yes, on Wan-AI's Hugging Face page |
How to Use the Wan 2.5 API on Kie.ai
Step 1. Get Your API Key
Step 2. Test Your Prompt in the Playground
Step 3. Create a Task and Get the Result
Tips Before You Ship
Describe the Sound in Your Prompt
The example prompts in the Kie.ai docs end with an "Ambient audio:" line that lists the sounds to generate, and every video in our tests came back with an audio track. Write the sounds you want the same way.
Expect Text in the Start Image to Change
In our chalkboard test the prices on the board disappeared within a few seconds, although the prompt asked for the lettering to stay still. If the start image contains text that must survive, check every frame or add the text after generation.
Pass a Public Image URL
image_url must be a publicly accessible URL, not file content. Files that are not hosted yet can be uploaded first with the Kie.ai File Upload API (Base64, file stream or remote URL).
Why Use Wan 2.5 on Kie.ai
Same Key Across the Kie.ai Market
All models in the Kie.ai market use the same authentication method: an Authorization: Bearer YOUR_API_KEY header with the key from your Kie.ai API Key Management page.
Free Credits to Start
New Kie.ai users receive 80 free credits for testing. Your account has one credit balance that all generation services draw from; check it with the Get Remaining Credits API.
Nothing to Host or Install
Kie.ai provides this model as an API with a playground; there is no app to download or install. There are no GPUs or model weights to manage.
FAQs About Alibaba Wan 2.5 Preview API
Not on Kie.ai: each video uses credits, as shown in the pricing card above. New Kie.ai users receive 80 free credits for testing.
There is no official Wan 2.5 download: as of October 2026, Wan-AI's Hugging Face page lists open weights for Wan 2.1 and Wan 2.2 but not Wan 2.5, and LMArena lists Wan 2.5 Preview as proprietary. Kie.ai provides this model as an API with a playground; there is no app to download or install.
Pricing for each resolution and duration is shown in the pricing card above.
It depends on what you need: Wan 2.5 Preview generates audio with the video, while Wan 2.2 has open weights you can host yourself. The comparison table above shows how they differ.
Wan 2.5 Preview is Alibaba's video model for text-to-video and image-to-video, with audio generated together with the video and clips up to 10 seconds long. On Kie.ai it runs as wan/2-5-text-to-video and wan/2-5-image-to-video.
Yes. Alibaba describes synchronized audio generation, and each of our four test videos on Kie.ai in October 2026 came with a non-silent audio track. Describe the sounds you want in the prompt; we could not check that every sound matched the prompt.
720p or 1080p, and 5 or 10 seconds per video. The parameter table above lists the output sizes we measured.
Yes, with wan/2-5-image-to-video: pass a publicly accessible image URL as the first frame along with your prompt. The robot and chalkboard examples above were made this way.
Yes. The Kie.ai docs for text-to-video say prompts support Chinese and English.
Create one on the API Key Management page after you sign up on Kie.ai. The steps above show where the key goes in a request.