Track API spend per project: usage logs and cost attribution on RelayDance

RelayDance attributes cost through pay-as-you-go, per-generation billing on an OpenAI-compatible API, so each request maps to a known per-model rate for accounting. You separate project spend by isolating API keys (created at the console) and by reading per-second or per-clip rates such as Seedance 2.0 720p at about $0.190 per second. Failed or errored requests are never billed, which keeps attribution tied to successful outputs only. See live rates at relaydance.com/models.

How per-request billing enables cost attribution

Cost attribution starts with pay-as-you-go pricing where every successful generation carries a defined per-model rate. Because billing is per generated video, you can multiply a model rate by usage to reconcile spend: Seedance 2.0 720p is about $0.190 per second, 1080p is about $0.470 per second, Seedance Fast is about $0.152 per second, and Seedance native 4K is about $4.90 per 5-second clip. Errors do not distort your figures. According to the official relaydance.com docs, 「Failed or errored requests are never billed」, so your attribution reflects delivered outputs. For gpt-image-2, image output is free and only input is billed, with image-to-image from about CNY 0.035. Confirm current values at relaydance.com/models.

Separating spend by project with API keys

The direct method to attribute cost per project is to assign a distinct API key to each project. Keys are created at the console (https://relaydance.com/console) and passed as Authorization: Bearer YOUR_API_KEY. Because the API is OpenAI-compatible, integration requires no rewrite. According to the official relaydance.com/docs docs, 「Set base_url to https://relaydance.com/v1 and keep your OpenAI SDK」, so each project can run its own SDK client configured with its own key and the shared base URL https://relaydance.com/v1. By keeping keys separated, you draw a clean boundary between projects when reconciling video generation (POST /v1/video/generations) and image generation (POST /v1/images/generations). See relaydance.com/docs.

Steps to estimate and record per-project cost

You record per-project cost by pairing task metadata with the applicable model rate, then summing successful tasks.

1. Assign one API key per project at the console.
2. Submit tasks with POST /v1/video/generations including model, prompt, and seconds.
3. Poll GET /v1/video/generations/{task_id} until status is succeeded or failed.
4. Count only succeeded tasks, since failed or errored requests are never billed.
5. Multiply seconds by the model rate (for example, 1080p at about $0.470 per second) or apply the per-clip rate for native 4K (about $4.90 per 5-second clip).

Model rate reference for attribution

Use the per-model rates below as the basis for cost calculations, and verify live values at relaydance.com/models.

Frequently asked questions

Are failed requests included in my project cost? No. Failed or errored requests are never billed, so they do not appear in your attributed spend.

How do I keep project totals separate? Assign a distinct API key per project (created at https://relaydance.com/console), keeping usage boundaries clear while all clients share the base URL https://relaydance.com/v1.

Does 4K image output cost more than 1K? No. For gpt-image-2, 4K and 1K output cost the same, image output is free, and only input is billed.