Image-to-video generation animates a still image into a short video clip. The image usually becomes the first frame, and your prompt describes the motion and the camera movement, so the subject, framing and style of the picture carry into the video.
Image-to-video is a mode of video generation. Instead of creating a scene from text alone, the model starts from a picture you supply and animates it. On most models the image becomes the first frame of the clip, so its subject, framing, colors and style are where the video begins. The prompt then describes what happens next: how the subject moves, what changes in the scene and how the camera travels, for example slowly zoom out and pan right.
This gives you more control than text-to-video. You can design or approve the exact look first, often with an image generation API, and then add motion to it.
Image-to-video uses the same endpoint as other video generation, POST /v1/videos/generations. Pass the image as a public URL in image, add a prompt for the motion, and set duration and resolution. On models that take a last frame, the second image goes in its own field, such as image_end. Many models choose the mode from what you attach, and a mode parameter selects it explicitly.
The output usually follows the aspect ratio of your image, so crop the image to the shape you want, such as 9:16 for vertical video, before you send it. Each model page lists the image formats and sizes it accepts. The request runs as an asynchronous job: the API returns a job_id at once, and you poll GET /v1/jobs/{job_id} until the video is ready.
Test candidates with your own image, because models differ in how faithfully they keep faces, products and printed text while adding motion. Also compare first and last frame support, maximum duration and resolution, native audio and generation time. Side-by-side model comparisons help narrow the list.
Image-to-video is billed like other video generation: most models charge per second of output at a rate set by the resolution, and many use the same rate for text-to-video and image-to-video. Some models also bill input images. Browse the video models, try one in the Playground, and read the video generation reference before you call the API with your key.
Pay as you go, or choose a plan with weekly allowances. Current rates are on the pricing page.
20 models, cheapest first where they share a billing unit. Each model's page has its full rate card, parameters and code.
Endpoints
/v1/videos/generationsPriced per second
| Model | Lowest rate | Applies to |
|---|---|---|
| Wan 2.2 Alibaba Cloud | $0.015 per second | Flash 480P |
| PixVerse V6 Max PixVerse | $0.030667$0.024534 per second | Standard 360p no audio |
| MiniMax H3 Lite MiniMax | $0.025 per second | 480p |
| Pixverse Avatar PixVerse | $0.025556 per second | 360p |
| Pixverse v6 PixVerse | $0.025556 per second | Standard 360p no audio |
| Pixverse C1 PixVerse | $0.030667 per second | Standard 360p no audio |
| Wan 2.1 Alibaba Cloud | $0.036 per second | Turbo |
| Wan 2.5 Alibaba Cloud | $0.05 per second | 480P |
| Model | Lowest rate | Applies to |
|---|---|---|
| MiniMax H3 Highspeed MiniMax | $0.06 per second | 768p |
| Seedance 2.0 Mini ByteDance | $0.070 per second | T2V/I2V 480P |
| Wan 3.0 Alibaba Cloud | $0.07 per second | Standard 480P |
| Kling 4.0 Flash Kling AI | $0.094 per second | Standard 720p |
| Grok Imagine Video 1.5 xAI | $0.096 per second | 480p |
| Seedance 2.5 Flash ByteDance | $0.135 per second | T2V/I2V 480P |
| HappyHorse 1.1 Alibaba Cloud | $0.14 per second | 720p |
| Kling 4.0 Kling AI | $0.157 per second | Standard 720p |
| Kling V3 Kling AI | $0.168 per second | Standard T2V/I2V |
| Kling 3.0 Turbo Kling AI | $0.18 per second | 720p |
| MiniMax H3 MiniMax | $0.18 per second | 768p |
| Seedance 2.5 ByteDance | $0.206 per second | T2V/I2V 480P |
Text-to-video creates the scene from a prompt alone. Image-to-video starts from a picture you supply, usually as the first frame, and the prompt describes the motion.
On most models, yes: image-to-video follows the shape of the source image. Crop the image to the ratio you need, such as 9:16 or 16:9, before you send it.
Yes, on models that support it. You send a second image as the last frame, and the model generates the motion between the two images.
Most models bill per second of generated video at a rate set by the resolution, often the same rate as text-to-video. Some models also bill input images. On EmpirioLabs AI you can pay as you go or choose a plan with weekly allowances.
Check out our pricing or reach out if you want your own model deployed on our stack.