Longer prompts and higher frame counts increase generation time. Start with default settings for best results.
| Model | Provider | Max Frames | Default Steps | Description |
|---|---|---|---|---|
| Wan 2.1 1.3B | Free | 81 | 20 | Smallest Wan model, fast generation |
| Wan 2.1 14B | Needs Credits | 81 | 20 | High quality video, needs fal-ai credits |
| HunyuanVideo | Needs Credits | 129 | 25 | Tencent strong video model, needs fal-ai credits |
| LTX Video | Needs Credits | 129 | 25 | High fidelity motion, strong prompt adherence |
| Mochi 1 Preview | Needs Credits | 129 | 25 | Genmo video model, needs fal-ai credits |
| CogVideoX 5B | Free | 49 | 25 | Zhipu AI video model |
| CogVideoX 2B | Free | 49 | 25 | Smaller CogVideoX, faster inference |
Send a POST request to /api/generate.php with a JSON body. The response is a JSON object with video metadata and a URL.
curl -X POST https://video.kim8.s4s.host/api/generate.php \
-H "Content-Type: application/json" \
-d '{
"prompt": "A cat playing piano in a jazz bar",
"model": "Wan-AI/Wan2.1-T2V-1.3B",
"num_frames": 33,
"num_inference_steps": 20,
"guidance_scale": 7.5,
"negative_prompt": "blurry, low quality",
"seed": 42
}'
Success response: {"success":true,"video_url":"...","filename":"...","size_mb":...,"time_ms":...}
Error response: {"success":false,"error":"...","hint":"..."}
Models powered by HuggingFace's free inference API. No credits required. Available models: Wan 2.1 1.3B, CogVideoX 5B, CogVideoX 2B. May have rate limits and longer queue times.
Models powered by fal-ai via HuggingFace router. Requires pre-paid credits on your HuggingFace account. Higher quality and faster generation. Models: Wan 2.1 14B, HunyuanVideo, LTX Video, Mochi 1.
## Video Generation API
Endpoint: POST /api/generate.php
Content-Type: application/json
### Parameters
- prompt (string, required): Video description
- model (string, required): Model ID from available models
- num_frames (int, optional): 1 to model max, default varies by model
- num_inference_steps (int, optional): 1 to 50, default varies by model
- guidance_scale (float, optional): 1 to 20, default varies by model
- negative_prompt (string, optional): What to avoid in the video
- seed (int, optional): Reproducible generation (omit for random)
### Response (success)
{
"success": true,
"video_url": "https://host/videos/filename.mp4",
"filename": "video_20250101_120000_abc12345.mp4",
"size_bytes": 1234567,
"size_mb": 1.18,
"model": "Wan-AI/Wan2.1-T2V-1.3B",
"provider": "hf-inference",
"num_frames": 33,
"time_ms": 45000
}
### Response (error)
{
"success": false,
"error": "Error message",
"hint": "Helpful suggestion",
"provider": "fal-ai",
"model": "...",
"http_code": 402
}
### Available Models
- Wan-AI/Wan2.1-T2V-1.3B (hf-inference, max 81 frames)
- Wan-AI/Wan2.1-T2V-14B (fal-ai, max 81 frames)
- tencent/HunyuanVideo (fal-ai, max 129 frames)
- Lightricks/LTX-Video (fal-ai, max 129 frames)
- genmo/mochi-1-preview (fal-ai, max 129 frames)
- zai-org/CogVideoX-5b (hf-inference, max 49 frames)
- zai-org/CogVideoX-2b (hf-inference, max 49 frames)
### Notes
- Generation takes 1-5 minutes. Use long timeouts.
- hf-inference models are free, fal-ai models need credits.
- video_url is a direct MP4 link valid for download.Central hub for all AI services and tools
Unified API gateway with no authentication required
AI-powered image analysis with multiple model engines
Generate, edit, and upscale images with AI
Browse 5700+ models by category