Öffentliche API

Videos und Bilder mit Ihren eigenen Tools generieren

Erstellen Sie einen Schlüssel, rufen Sie die API mit Bearer-Authentifizierung auf und fragen Sie den Aufgabenstatus ab, bis Ihr Ergebnis bereit ist. Funktioniert mit Claude Code, Codex und jedem HTTP-Client.

Schlüssel verwalten

Auf dieser Seite

Schnellstart

1
Schlüssel erstellen
Generieren Sie einen API-Key in Ihrem Dashboard. Er wird nur einmal angezeigt, bewahren Sie ihn daher sicher auf.
Schlüssel verwalten

API aufrufen

Senden Sie eine POST-Anfrage mit Ihrem Bearer-Schlüssel und einem Idempotency-Key-Header.

curl -X POST https://www.seedance2ai.io/api/v1/video/seedance2 \
  -H "Authorization: Bearer $SEEDANCE_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: demo-video-001" \
  -d '{
    "mode": "text-to-video",
    "quality_tier": "standard",
    "prompt": "A cinematic shot of a glass train crossing a snowy mountain bridge",
    "aspect_ratio": "16:9",
    "duration": "5",
    "resolution": "720p"
  }'

Ergebnis abrufen

Fragen Sie den Task-Status-Endpunkt ab, bis der Status „abgeschlossen“ oder „fehlgeschlagen“ ist, und lesen Sie dann die Ausgabe.

while true; do
  TASK=$(curl -s https://www.seedance2ai.io/api/v1/tasks/sd2_xxxxx \
    -H "Authorization: Bearer $SEEDANCE_API_KEY")
  STATUS=$(echo "$TASK" | jq -r '.status')
  if [ "$STATUS" = "completed" ] || [ "$STATUS" = "failed" ]; then
    echo "$TASK" | jq
    break
  fi
  sleep 10
done

Authentifizierung

Jede Anfrage wird mit einem Bearer-API-Schlüssel im Authorization-Header authentifiziert.

Base URL: https://www.seedance2ai.io/api/v1
Idempotency-Key: Send a unique value with every POST. Retrying with the same key and body within 24 hours returns the original task instead of charging for a second one.
Credits: Calls spend credits from the key owner's personal balance. Team credits are not used.
Media URLs: Reference images, videos and audio must be public HTTPS URLs that can be fetched without cookies or a login.

http

Authorization: Bearer sk_live_your_api_key
Content-Type: application/json
Idempotency-Key: a-unique-id-per-request

Seedance 2.0 Video

Text, Bild oder Medien zu Video.

Modes: Text, image, reference media
Duration: 4 to 15 seconds
Quality tiers: mini, standard, pro
Task ID: sd2_…

Auftrag einreichen

POST

/api/v1/video/seedance2

Parameter

modestringOptionalStandard text-to-video
Generation mode.
text-to-videoimage-to-videomedia-to-video
quality_tierstringOptionalStandard standard
Quality tier.
ministandardpro
channelstringOptionalStandard standard
Rendering channel.
standardrealwild
promptstringErforderlich
Text prompt.
3–10000 chars
aspect_ratiostringOptionalStandard 16:9
Aspect ratio.
1:121:94:33:416:99:16adaptive
durationstringOptionalStandard 5
Clip duration.
4–15 (seconds)
resolutionstringOptionalStandard 720p
Output resolution. 1080p-plus and 4k are available on pro only.
720p1080p1080p-plus4k
image_urlstring (URL)Optional
Start frame — required for image-to-video.
public https URL, ≤ 30 MB
end_image_urlstring (URL)Optional
Optional end frame.
public https URL, ≤ 30 MB
media_urlsstring[] (URL)Optional
Required for media-to-video. Images are limited to 30 MB; video URLs must support duration probing.
≤ 12 public https image/video URLs
generate_audiobooleanOptionalStandard true
Generate an audio track.
truefalse
fixed_lensbooleanOptionalStandard false
Lock the camera lens.
truefalse
seedintegerOptional
Optional deterministic seed.
-1 – 4294967295

Seedance prompt guide

Seedance prompts work best as concrete director instructions. Describe the subject, action, scene, camera motion, visual style, and constraints, then map every reference asset to the right API parameter.

Use image_url for the opening frame, end_image_url for the closing frame, and media_urls for character, motion, style, source-video, or extension references.
Describe asset roles in natural language inside prompt. Do not send raw asset ids, asset:// URLs, or in-app-only labels such as @图片1.
For edits say “strictly edit the supplied source video”; for extensions say “continue from the supplied previous clip.”
For complex scenes, use ordered shots plus a style and constraint package: stable faces, natural motion, no deformation, no watermark, no logo, and no subtitles unless intended.

Anfrage

curl -X POST https://www.seedance2ai.io/api/v1/video/seedance2 \
  -H "Authorization: Bearer $SEEDANCE_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: demo-video-001" \
  -d '{
    "mode": "text-to-video",
    "quality_tier": "standard",
    "prompt": "A cinematic shot of a glass train crossing a snowy mountain bridge",
    "aspect_ratio": "16:9",
    "duration": "5",
    "resolution": "720p"
  }'

Antwort202

json

{
  "id": "sd2_xxxxx",
  "status": "processing",
  "model": "seedance2",
  "quality_tier": "standard",
  "channel": "standard",
  "credits_used": 30
}

Die in diesen Beispielen gezeigten credits_used-Werte dienen nur zur Veranschaulichung. Der tatsächliche Verbrauch wird beim Absenden berechnet und von der API zurückgegeben.

Aufgabenstatus prüfen

GET

/api/v1/tasks/{id}

status

processing: Accepted and running. Keep polling.
completed: Finished. Read output, or result_urls for a flat list of files.
failed: Ended without a result. error says why and credits_refunded shows what was returned.

Poll every 10 seconds or so and stop at completed or failed. Status reads have their own rate limit, so polling never uses up your submit budget.

Anfrage

curl https://www.seedance2ai.io/api/v1/tasks/sd2_xxxxx \
  -H "Authorization: Bearer $SEEDANCE_API_KEY"

Antwort200

json

{
  "id": "sd2_xxxxx",
  "type": "video",
  "status": "completed",
  "model": "seedance2",
  "quality_tier": "standard",
  "channel": "standard",
  "resolution": "720p",
  "credits_used": 30,
  "credits_refunded": 0,
  "output": {
    "video_url": "https://...",
    "last_frame_url": "https://...",
    "seed": 123
  },
  "result_urls": ["https://..."],
  "error": null,
  "created_at": "2026-06-03T10:00:00.000Z",
  "updated_at": "2026-06-03T10:03:12.000Z"
}

Wait for the result

while true; do
  TASK=$(curl -s https://www.seedance2ai.io/api/v1/tasks/sd2_xxxxx \
    -H "Authorization: Bearer $SEEDANCE_API_KEY")
  STATUS=$(echo "$TASK" | jq -r '.status')
  if [ "$STATUS" = "completed" ] || [ "$STATUS" = "failed" ]; then
    echo "$TASK" | jq
    break
  fi
  sleep 10
done

Tasks and balance

Read endpoints for building a task history, showing remaining credits, or discovering models at runtime. They share the read rate limit.

List tasks

GET

/api/v1/tasks

Returns the tasks this account created through the API in the last 30 days, newest first. Each item has the same shape as a single task. Use limit (1 to 50, default 20) and pass next_cursor as before to fetch the next page.

curl "https://www.seedance2ai.io/api/v1/tasks?limit=20" \
  -H "Authorization: Bearer $SEEDANCE_API_KEY"

json

{
  "data": [
    {
      "id": "sd2_xxxxx",
      "type": "video",
      "status": "completed",
      "model": "seedance2",
      "credits_used": 30,
      "result_urls": ["https://..."]
    }
  ],
  "has_more": false,
  "next_cursor": null
}

Check balance

GET

/api/v1/balance

Returns the credits the key owner can spend through the API. subscription_credits is the part that came from a plan and expires at subscription_credits_expire_at.

curl "https://www.seedance2ai.io/api/v1/balance" \
  -H "Authorization: Bearer $SEEDANCE_API_KEY"

json

{
  "credits": 1200,
  "subscription_credits": 1000,
  "subscription_credits_expire_at": "2026-07-01T00:00:00.000Z"
}

List models

GET

/api/v1/models

Returns every model the API serves with its endpoint, accepted modes and task ID prefix. Useful for building a model picker without hard-coding this page.

curl "https://www.seedance2ai.io/api/v1/models" \
  -H "Authorization: Bearer $SEEDANCE_API_KEY"

json

{
  "data": [
    {
      "id": "seedance2",
      "name": "Seedance 2.0",
      "type": "video",
      "method": "POST",
      "endpoint": "/api/v1/video/seedance2",
      "modes": ["text-to-video", "image-to-video", "media-to-video"],
      "task_id_prefix": "sd2_"
    }
  ]
}

Fehlercodes

Every error has the same shape. Branch on code, and show message to a person. A task that was accepted can still fail later, so also check status when you poll.

json

{
  "error": {
    "code": "invalid_request",
    "message": "Invalid request body: prompt: String must contain at least 3 character(s)"
  }
}

Code	HTTP	Bedeutung
`unauthorized`	`401`	Missing, invalid, or revoked API key.
`invalid_request`	`400`	Bad input or unsupported field. The message names the first field that failed; details lists them all.
`purchase_required`	`403`	Seedance 2.5 requires a successful credits or plan purchase; free promotional credits alone do not unlock it. The GPT Image 2.5 economy channel has the same requirement.
`insufficient_credits`	`402`	Not enough credits. Seedance 2.5 always returns HTTP 402 with insufficient_credits; when an authoritative quote is available, it may also include required_credits, available_credits, pricing_version, and promotion_applied.
`rate_limited`	`429`	Too many requests for this account. See Rate limits below. Seedance requests that validate reference media are limited to 3 per 60 seconds per account and source IP, with a project-wide safety cap. Retry after the Retry-After response header (in seconds).
`idempotency_conflict`	`409`	Same Idempotency-Key reused with a different body, or still running.
`service_busy`	`503`	Temporary upstream or credit-concurrency issue. Retry.
`not_found`	`404`	Task does not exist or does not belong to this key owner.
`internal_error`	`500`	Unexpected server-side failure.

Rate limits

A request over the limit returns HTTP 429 with a Retry-After header in seconds, plus X-RateLimit-Limit and X-RateLimit-Reset. Wait that long, then retry with the same Idempotency-Key.

Submit a job: 30 per 60 seconds; Every POST endpoint, counted per account across all of its keys.
Read: 120 per 60 seconds; Task status, task list, balance and models. Counted separately, so polling never uses up the submit budget.
Reference media validation: 3 per 60 seconds; Seedance requests that include reference media, per account and source IP.

Mit Claude Code verwenden

Mit Copy for AI exportieren Sie die vollständige API als sauberes Markdown, fügen es in Claude Code oder Codex ein und lassen den Agenten die Integration einrichten. Oder Sie verweisen den Agenten direkt auf den Roh-Text-Endpunkt.

llms-api.txt öffnen DeepSeek Harness plugin

Skill

Installieren Sie die Seedance-Fertigkeit in Claude Code, Codex oder einen beliebigen Agenten – sie schreibt filmreife Prompts und rendert diese automatisch über die API für Sie.

bash

# Download the one-file skill from Seedance 2.0 AI
mkdir -p ~/.claude/skills/seedance
curl --fail --location -H "Authorization: Bearer $SEEDANCE_API_KEY" https://www.seedance2ai.io/downloads/seedance-skill.md -o ~/.claude/skills/seedance/SKILL.md.tmp &&
mv ~/.claude/skills/seedance/SKILL.md.tmp ~/.claude/skills/seedance/SKILL.md

Skill herunterladen

FAQ

Wer kann die API nutzen?: Jeder mit Credits. Es gibt keine Planbeschränkung – jedes Konto mit Credits kann einen Schlüssel erstellen und die API aufrufen.
Wie bekomme ich Credits?: Laden Sie Guthaben auf oder abonnieren Sie auf der Preisseite. Guthaben wird Ihrem Konto sofort gutgeschrieben.
Werden Credits zwischen der Website und der API geteilt?: Ja – Ihr persönliches Guthaben wird sowohl für die Web-Generierung als auch für Ihre API-Aufrufe verwendet. In einem Team verwenden API-Aufrufe das persönliche Guthaben jedes Mitglieds, nicht das gemeinsame Team-Guthaben.
Wie starte ich?: Erstellen Sie einen API-Key in Ihrem Dashboard und rufen Sie dann die Endpoints mit einem Bearer-Token auf (siehe Quickstart oben).
Was passiert, wenn ein Aufruf `insufficient_credits` zurückgibt?: Ihr Guthaben ist leer – laden Sie auf der Preisseite auf und versuchen Sie es erneut.

Videos und Bilder mit Ihren eigenen Tools generieren

Schnellstart

Schlüssel erstellen

API aufrufen

Ergebnis abrufen

Authentifizierung

Seedance 2.0 Video

Auftrag einreichen

Parameter

Seedance prompt guide

Anfrage

Antwort202

Aufgabenstatus prüfen

status

Anfrage

Antwort200

Wait for the result

Tasks and balance

List tasks

Check balance

List models

Fehlercodes

Rate limits

Mit Claude Code verwenden

Skill

FAQ