Pixmind

Published API route

Seedance 2.5 Flash API

Add 480p/720p video creation, multimodal references and source-video workflows to your application. Submit one request, retain its task ID, and query the result through PixMind's unified API.

Dark editing workstation with several video-frame previews for asynchronous creative jobs
Original AI-created prompt concept, separate from the Flash video sample.
480p / 720p4–30 seconds per clip10 images · 5 videos · 5 audio clipsGenerated audio on or off

Submit a video request

Use the exact model ID and your own PixMind API key. This is the PixMind endpoint and its camelCase request schema.

curl --request POST 'https://aihub-admin.aimix.pro/api-platform/v1/generations' \
  --header 'Authorization: Bearer YOUR_PIXMIND_API_KEY' \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "seedance-2-5-flash",
  "prompt": "One uninterrupted low tracking shot inside a botanical greenhouse at dawn. A crane walks past a stone lantern and gently spreads its wings beside a shallow pool. The camera tracks sideways at a steady speed. Soft mist, warm sunlight, realistic feathers, no other birds and no cuts.",
  "duration": 8,
  "resolution": "720p",
  "aspectRatio": "16:9",
  "generateAudio": false
}'

Retain the task ID and query the result

The creation response contains a task ID. Query it until completion or failure, then save the returned video URL. A delayed response is a reason to query the existing ID, not submit another paid job.

curl 'https://aihub-admin.aimix.pro/api-platform/v1/tasks/YOUR_TASK_ID' \
  --header 'Authorization: Bearer YOUR_PIXMIND_API_KEY'

Confirmed PixMind credit prices

API billing is separate from the Studio wallet. Confirmed rates are 68 credits/second at 480p and 140 credits/second at 720p, exactly 80% of the Studio rates.

ResolutionAPI credits / secondStudio credits / second
480p6885
720p140175

Automatic length first reserves 30 output seconds. After completion, the actual output length is rounded upward to the next whole second and unused credits are released. Editing uses automatic length.

Flash API integration questions

Which model ID should my application send?

Send seedance-2-5-flash. The upstream alias doubao-seedance-2-5-flash is normalized to this model. Do not send seedance-2.5-flash, and do not substitute the separate seedance-2.5 model.

Which endpoint and request schema does PixMind use?

Use POST /api-platform/v1/generations with model, prompt, duration, resolution and aspectRatio. PixMind also accepts referenceImageUrls, referenceVideoUrls, referenceAudioUrls and the firstFrameUrl/lastFrameUrl fields. The upstream APIMart schema is translated by the adapter.

Which resolutions and durations can I request?

Use 480p or 720p and an integer duration from 4 to 30 seconds. Set duration to -1 for automatic output duration. Editing requires -1 or an omitted duration and follows the source video. Unsupported resolutions are rejected before generation.

What are the reference limits?

The combined image limit is 10, with up to 5 videos and 5 audio clips. Each video or audio clip is 2–30 seconds; totals cannot exceed 30 seconds per media type. Editing source videos must be 4–30 seconds. Use public media URLs or approved asset references.

How should I submit frames, editing and extension?

Use firstFrameUrl and optionally lastFrameUrl for frame-guided work. Set mode to edit or extend and provide a source video for those workflows. Frame-only, edit and extend requests require aspectRatio adaptive. Draft fields are not supported.

How are API credits calculated?

The confirmed output rates are 68 credits per second at 480p and 140 at 720p. Eight seconds costs 544 or 1120 credits. Automatic duration reserves 30 seconds, then settles actual output duration rounded up to whole seconds. The API wallet and its ledger are separate from Studio.

When is the result ready to download?

Keep querying the original task ID while it is pending or processing. Completion makes the video URL available. Flash can briefly wait for supplier usage data after rendering. If returnLastFrame is true, the completed response also exposes a saved lastFrameUrl.

Should I retry creation after a timeout?

Query the saved task ID first. A creation timeout can mean the supplier accepted the job; automatically submitting again can create duplicate work. If no queryable ID was returned, reconcile the supplier record before resubmitting. Submitted Flash jobs do not support cancellation.

What happens to credits when generation fails?

Rejected inputs do not start generation. A confirmed generation failure returns the original fixed charge or releases an automatic-duration reservation. Polling or storage errors keep the original job under reconciliation rather than issuing a refund while it may still run.