The pipeline behind our series, as an API.
One key, one job model. The same video generation that renders our own episodes — multi-model, async, with failed jobs refunded automatically.
Billed in credits against your SeedAgent balance · Failed renders refunded automatically
Three endpoints and a webhook.
Submit and get a job id back. Collect the result by webhook or by polling. Video generation only in this version — image, voice and music are not part of the API yet.
/v1/modelsThe catalogue your key may generate with: capabilities, hard constraints and the credit rate for each model. Read it instead of hard-coding limits.
/v1/videosText-to-video, image-to-video, first/last-frame interpolation and reference images. Returns a job id and the exact price before anything renders.
/v1/videos/{id}Four states — queued, processing, completed, failed. Scoped to the calling key, so one partner's keys never read another's jobs.
video.completedRegister a webhook_url and we POST the finished job, HMAC-signed with a replay-proof timestamp. Polling stays available for reconciliation.
Plain HTTP. No SDK to install.
Flat JSON in, a job id out. Every field is validated against the model's published constraints, so a request that respected GET /v1/models is never rejected later.
curl -X POST https://seedagent.ai/api/v1/videos \
-H "Authorization: Bearer $SEEDAGENT_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "sa-video-pro",
"prompt": "A lone figure on a rain-slicked rooftop at dusk.",
"duration": 5,
"aspect_ratio": "16:9",
"resolution": "1080P",
"webhook_url": "https://yourapp.com/hooks/seedagent"
}'Full reference — every field, every error code, and webhook signature verification — is in the partner documentation we send with your key.
Async by default, because a good shot takes a minute.
Submitting never holds a connection open. What we guarantee about money and credentials is short enough to list in full.
Automatically, by the same pipeline that runs our own productions — not by a support ticket.
credits_charged is in the submit response, and the per-second rate is published per model.
A request that would overdraw is rejected up front. Nothing is queued and nothing is charged.
We store only a hash. Rotate on suspicion and the old secret dies instantly, keeping your key id and history.
Pay per second of finished output.
Billed in credits against your SeedAgent account balance — the same unit and the same rates the product itself charges, with volume terms agreed per partner.
- Per second of output, with a multiplier for 1080p. Each model publishes its own rate in GET /v1/models.
- The exact charge is returned with the job id, before rendering starts.
- Failed renders are refunded automatically, in full.
- Top up from your normal account billing page — no separate invoice to reconcile.
Tell us what you are building.
Keys are issued by hand during the beta. Leave your work email and what you have in mind, and we will come back with a key and terms sized to your volume.