Your agent has an idea. Give it a video API.
Make a DeepSeek whale girl short, animate an original character, or turn a comic scene into motion. Your agent directs the video. We provide the models and the API.
Character resources are free to read. Video generation uses your account's paid credits, priced by model, resolution and duration.
Tell your agent what you want to make
"Find references for DeepSeek whale girl, help me create a character image, then make a short video I can share with my friends."
An existing video made on this site. Open the example for its workflow and sources.
See the Feiyu Shake exampleConnect once. Keep creating in your agent.
Read the guide, add your key to your agent's local environment, and let it call the video API. You can use HTTP from a terminal or any agent that can make authenticated requests. No MCP installation is required.
Read the agent guideFrom an idea to a video in your workspace
- 1
Read the character references
Ask your agent to read the character profile and inspect the reference images. Names, visual details and source links help it understand which version you mean. You can also bring an original character. Use your existing image generation tool to make the image you want to animate.
- 2
Let your agent direct the video
Your agent chooses a model, writes the prompt and sets the aspect ratio, duration and resolution. It can upload a local reference image, use an existing HTTPS image URL, or prepare motion and audio references. The video API accepts native model inputs without adding a preset prompt.
- 3
Check the quote and generate
The quote endpoint returns the exact site credits for the request. Your agent checks the balance and submits the video with a unique idempotency key. Keeping that key on retries prevents duplicate jobs. Your existing paid account balance funds the video.
- 4
Download and share your work
Generation runs as a background job. Your agent polls the task, downloads the completed MP4 with the same API key, and saves it locally. Review the video before posting it to your community, X, Bilibili or TikTok. Publishing remains your choice.
Add video credits to your agent's workflow
Character references are free. Add paid credits when you want your agent to generate videos. Free image credits cannot be used for video.
The display currency follows your selected language. Amounts are estimates before tax and exchange fees. Checkout determines the payment currency and final total; Waffo's local conversion includes a 2–4% fee. WeChat Pay uses separate conversion.
Checkout with 
and 
7-day refunds for unused packsRefund policy
Choose the video model for the job
Use Mini for quick drafts, Fast for shorter iteration cycles, the standard model for more output resolutions, or Seedance 2.5 for clips up to 30 seconds. Text, image and reference inputs use the same model-specific API contract. Your agent can read the current model list without signing in.
| Model | Maximum length | 5-second video at 480p |
|---|---|---|
Seedance 2.5seedance-2-5 | 30s | 30 credits |
Seedance 2.0seedance-2 | 15s | 20 credits |
Seedance 2.0 Fastseedance-2-fast | 15s | 13 credits |
Seedance 2.0 Miniseedance-2-mini | 15s | 8 credits |
Examples use text-to-video with no reference video. Charges are rounded up for the whole request. Reference-video pricing includes the output duration plus the measured reference-video duration. Query the quote endpoint for the exact amount before submitting.
The API uses the same paid credits as the website. Free image allowances do not fund videos. Successful videos settle the reserved credits after saving. Confirmed generation failures release the reservation. A submission whose outcome is unknown stays under review to avoid paying twice.
Give your agent the character context
AI gijinka turns an AI model into a character. DeepSeek whale girl, also called 大肥鱼, is one example. The character reference library links to specific designs, creators and works, so your agent can distinguish a visual reference from a claim about official branding. Read the source and licensing notes before reusing someone else's artwork.
DeepSeek whale girl
Start here for 大肥鱼: appearance, aliases, creators and a downloadable three-view PNG.
AI gijinka character library
Meet the characters for Claude, Gemini, Qwen and other AI models.
Community fanworks
Explore comics and character stories, with links to the original works.
Character creation ideas
See worked examples and reference-led image prompts for your own creations.
A video API your agent can read and use
The public Markdown guide contains the full request flow, upload commands, model IDs, polling states and download instructions. Give your agent its URL together with your creative request. Set the key privately in the environment rather than pasting it into the conversation. The guide itself is public and contains no account credentials.
POST /api/agent/v1/generations
Authorization: Bearer $GSA_API_KEY
Idempotency-Key: my-character-video-001
Content-Type: application/json
{
"model": "seedance-2-mini",
"endpoint": "text-to-video",
"provider": "seeapi",
"input": {
"prompt": "A blue-haired whale girl waves by the sea",
"duration_seconds": 5,
"resolution": "480p",
"aspect_ratio": "9:16",
"generate_audio": true
}
}Example request only. Submitting it with a valid key uses paid credits. The API forwards your native input to the selected model. It does not generate images or publish to social accounts.
Before you connect
You can keep using the agent and image tools you already have. This service supplies video generation through your Gender Swap AI account. A subscription to an agent does not include video credits here. Create a key, check the quote, and add paid credits when you want to generate.
Yes. The agent controls the prompt and supported model inputs. Use text-to-video for a written scene, image-to-video for a first frame, or reference-to-video for character, motion and audio references. Public HTTPS image URLs work directly. Upload local images, video and audio with the signed upload flow in the guide.
We measure uploaded media to check the model limits and calculate reference-video charges. Uploads go straight to private storage through a temporary signed URL. The video API does not proxy large files through the generation request. The guide lists file formats, sizes and the upload steps.
Reuse the same Idempotency-Key and request body to recover the original job. A changed request with the same key returns a conflict. Poll queued, processing and saving jobs until they finish. A confirmed failure releases reserved credits. If a provider submission is uncertain, the task enters review. Do not submit another paid job to recover it.
You manage your own key while signed in. The full secret is shown once and the server stores only its hash. You can replace or revoke it here. Your key can access your own account's video tasks and downloads. Inputs are shared with the generation provider for processing, and site video outputs are retained for 30 days. Save finished videos locally.
Yes. The first version is a native HTTP API with a public Markdown guide. An agent that can read a URL and make HTTP requests can use it. The key authorizes video API calls only. It cannot sign into your account, buy credit packs, change billing, or publish your content.
Keep the idea in your conversation. Get the video back there too.
Start with the guide and your personal API key. Your agent can take care of the request, the waiting and the download.
Get your API key