How to chain gpt-image-2 into Seedance: generate an image, then animate it

To chain gpt-image-2 into Seedance on RelayDance, call POST /v1/images/generations with model gpt-image-2 to create a still image. Then submit that image as reference media in metadata.content[] to POST /v1/video/generations, cite it in the prompt as @image1, and poll GET /v1/video/generations/{task_id} until status is succeeded. Both calls use the same OpenAI-compatible base URL, https://relaydance.com/v1, and the same API key.

What do you need before chaining the two models?

You need a RelayDance API key, the OpenAI SDK you already use, and the base URL https://relaydance.com/v1. According to the relaydance.com docs, “Set base_url to https://relaydance.com/v1 and keep your OpenAI SDK”, so no new client library is required. Create a key at https://relaydance.com/console and send it as Authorization: Bearer YOUR_API_KEY on every request. The image call and the video call both go through this one base URL, so the chain is two HTTP requests plus status polling. Accounts are funded with USDT or Stripe card, and no BytePlus enterprise account or KYC is required to reach Seedance. Endpoint and parameter details are in the RelayDance docs.

How do you turn a gpt-image-2 still into a Seedance clip?

You generate the image with gpt-image-2, pass it to Seedance as reference media, and retrieve the finished video.

1. Generate the still: send POST /v1/images/generations with model gpt-image-2 and a prompt describing the opening frame.
2. Submit the video task: send POST /v1/video/generations with model, prompt, seconds, and metadata. Place the generated image in metadata.content[] and reference it in the prompt as @image1. Set ratio, resolution and generate_audio in metadata as needed.
3. Track the task: poll GET /v1/video/generations/{task_id} until status is succeeded or failed, or set metadata.callback_url to receive the final state as a POST to your server.
4. Collect the result: a succeeded task contains the video url.

Each request accepts up to 9 reference images, 3 reference videos and 3 audio tracks, and clips run up to 15 seconds.

What does the image-to-video chain cost?

The video step is billed per second of generated video, while gpt-image-2 bills input and charges nothing for image output.

For example, a 10-second Seedance 2.0 clip costs about $1.90 at 720p and about $4.70 at 1080p. On the image side, 4K and 1K gpt-image-2 output cost the same, and image-to-image starts from about CNY 0.035. According to the relaydance.com docs, “Failed or errored requests are never billed”, so a task that ends in status failed adds no charge. Live rates are listed at relaydance.com/models.

FAQ

The questions below cover common points when chaining gpt-image-2 and Seedance on RelayDance.

Can I animate several gpt-image-2 images in one video request? Yes. A request accepts up to 9 reference images in metadata.content[], cited in the prompt as @image1 to @image9.

How long can the animated clip be? Clips run up to 15 seconds, set with the seconds parameter on POST /v1/video/generations.

Do I have to poll for the video result? No. Set metadata.callback_url and RelayDance POSTs the final task state to your server. Otherwise, poll GET /v1/video/generations/{task_id} until status is succeeded or failed.