GPT Image 2.5 is live — OpenAI's newest image model, targeted edits that leave the rest of the frame alone
rreAPI Docs

seedance-2-5-flash

Generate 4–30 second videos at 480p or 720p from text, frames and mixed references, or edit a source video with the native content array.

Seedance 2.5 Flash supports text-to-video, first/last-frame guidance, mixed references and video editing. See the model page for the playground and current pricing.

Quick example

curl https://reapi.ai/api/v1/videos/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2-5-flash",
    "content": [{"type": "text", "text": "A white ceramic cup slowly rotates under warm studio lighting."}],
    "resolution": "480p",
    "duration": 4,
    "ratio": "16:9",
    "generate_audio": false
  }'
import requests

response = requests.post(
    "https://reapi.ai/api/v1/videos/generations",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={
        "model": "seedance-2-5-flash",
        "content": [{"type": "text", "text": "A white ceramic cup slowly rotates under warm studio lighting."}],
        "resolution": "480p",
        "duration": 4,
        "ratio": "16:9",
        "generate_audio": False,
    },
    timeout=60,
)
response.raise_for_status()
task_id = response.json()["id"]
const response = await fetch('https://reapi.ai/api/v1/videos/generations', {
  method: 'POST',
  headers: {
    Authorization: 'Bearer YOUR_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'seedance-2-5-flash',
    content: [{ type: 'text', text: 'A white ceramic cup slowly rotates under warm studio lighting.' }],
    resolution: '480p', duration: 4, ratio: '16:9', generate_audio: false,
  }),
});
if (!response.ok) throw new Error(`Request failed: ${response.status}`);
const { id: taskId } = await response.json();
package main

import (
    "encoding/json"
    "fmt"
    "net/http"
    "strings"
    "time"
)

func main() {
    body := strings.NewReader(`{
      "model": "seedance-2-5-flash",
      "content": [{"type": "text", "text": "A white ceramic cup slowly rotates under warm studio lighting."}],
      "resolution": "480p", "duration": 4,
      "ratio": "16:9", "generate_audio": false
    }`)
    req, err := http.NewRequest("POST", "https://reapi.ai/api/v1/videos/generations", body)
    if err != nil { panic(err) }
    req.Header.Set("Authorization", "Bearer YOUR_API_KEY")
    req.Header.Set("Content-Type", "application/json")
    res, err := (&http.Client{Timeout: 60 * time.Second}).Do(req)
    if err != nil { panic(err) }
    defer res.Body.Close()
    if res.StatusCode >= 300 { panic(res.Status) }
    var task struct { ID string `json:"id"` }
    if err := json.NewDecoder(res.Body).Decode(&task); err != nil { panic(err) }
    fmt.Println(task.ID)
}

Endpoint

POST /api/v1/videos/generations, with Authorization: Bearer YOUR_API_KEY and Content-Type: application/json.

The response contains a reAPI task id. Poll GET /api/v1/tasks/:id with the same authorization until completed or failed. Polling does not create a new generation. Do not resubmit merely because generation takes longer than expected.

Parameters

FieldRequiredDefaultMeaning
modelYes—seedance-2-5-flash
contentYes—Array containing a text block and optional media blocks described below
resolutionNo720p480p or 720p
durationNo5 for generationInteger 4–30 for generation; omit or use -1 for editing
ratioNo16:9 for text/reference; adaptive for frames/editing16:9, 9:16, 1:1, 4:3, 3:4, 21:9; frame guidance and editing also accept adaptive. Editing preserves the source ratio
generate_audioNotrueGenerate an audio track; false produces silent video
omni_reference_task_typeNogenerategenerate or edit
content_filterNotrueOnly true is supported

This model uses native content[], not top-level prompt, image_urls or video_urls. Omit optional fields to use their defaults; explicit null is not an omitted field. Model parameters, including extensions, are forwarded for model-side validation. Unsupported parameters can fail the task; acceptance alone does not prove an option affects output.

Content blocks

TypeShapeRoles
Text{"type":"text","text":"Describe the shot"}Text is required in every mode
Image{"type":"image_url","image_url":{"url":"https://example.com/image.jpg"},"role":"reference_image"}reference_image, first_frame, last_frame
Video{"type":"video_url","video_url":{"url":"https://example.com/video.mp4"},"role":"reference_video"}reference_video
Audio{"type":"audio_url","audio_url":{"url":"https://example.com/audio.mp3"},"role":"reference_audio"}reference_audio

Every media URL must be public HTTP(S). Base64 and data: inputs are not supported.

Modes and media requirements

  • Text: supply a text block with no media. Choose a fixed aspect ratio.
  • First/last frame: supply text and one first_frame; optionally add one last_frame. Do not mix frame roles with reference media. Omit ratio or use adaptive to follow the image.
  • Reference generation: supply text with up to 10 images, 5 videos and 5 audio clips. Describe what each reference should contribute. Choose a fixed aspect ratio.
  • Editing: supply text and exactly one reference video, set omni_reference_task_type: "edit", and omit duration or set -1. Omit ratio or use adaptive.

Images must be JPEG, PNG or WebP, at most 30 MB each, with each edge 300–6,000 pixels and width/height ratio 0.4–2.5. Videos must be MP4 or MOV, at most 200 MB each, 24–60 fps, with each edge 300–6,000 pixels and total pixels 409,600–8,295,044. Reference videos are 2–30 seconds each with at most 30 seconds combined; editing takes one 4–30 second video. Audio must be WAV or MP3, at most 15 MB each, 2–30 seconds each and at most 30 seconds combined.

Editing example:

{
  "model": "seedance-2-5-flash",
  "content": [
    {"type": "text", "text": "Change the white cup to deep blue. Preserve its movement, framing and lighting."},
    {"type": "video_url", "video_url": {"url": "https://example.com/source.mp4"}, "role": "reference_video"}
  ],
  "omni_reference_task_type": "edit",
  "resolution": "480p",
  "duration": -1,
  "ratio": "adaptive",
  "generate_audio": false
}

Draft/final workflows, video extension, 1080p, seed, watermark and last-frame output are not part of this model's documented surface.

Pricing dimensions and formula

Rates depend on resolution and whether the input contains video. Audio generation does not change the rate. Images and audio do not add billed seconds. See current rates.

Let output be the requested generation seconds, and input = ceil(sum of source video playback seconds - 0.000001):

  • Without video input: billed_seconds = output.
  • Reference generation with video: billed_seconds = output + max(output, input).
  • Editing: billed_seconds = 2 * input.
  • credits = ceil(rate_per_billed_second * billed_seconds * 1000).

One credit is $0.001. The server measures input videos before creating a task. If the duration cannot be determined, it returns PRICING_UNAVAILABLE without charging or submitting the task. Video playback duration can differ slightly from a rounded duration label; a source slightly longer than four seconds rounds up to five billed input seconds. An estimate based on browser metadata is provisional; server measurement is authoritative. Failed tasks refund the reserved credits.

Output

{
  "id": "task_example",
  "model": "seedance-2-5-flash",
  "status": "completed",
  "output": {"video_urls": ["https://example.com/result.mp4"]}
}

Use the reAPI task ID for polling. A successful response means the task was accepted; only the completed result confirms a video was produced.

Errors

CodeMeaning
10001 / 10003Missing / invalid API key
30001Insufficient credits
20001 / 20003Invalid JSON / parameter value
30002Pricing cannot be determined
80006Content review rejected the request
80007Model rejected the supplied parameters
80001 / 80003Submission / generation failed

Errors use { "error": { "code", "message", "request_id" } }. See the error catalog for the complete definitions and task failure envelope.

Start with one shot and one camera movement. In editing, state both the change and what must remain. Inspect the full result, including its ending and audio; a successful task is not a guarantee of exact visual fidelity.

Use standard Seedance 2.5 when you need its larger reference capacity or 1080p tier. Flash is a separate model, not its Draft mode.

Table of Contents