What kinds of inputs does Seedance 2.0 support for video generation?

Seedance 2.0 supports multimodal inputs including text prompts, up to 9 images, up to 3 short video clips, and up to 3 audio files, which can be freely combined for rich, controllable generation.

Can Seedance 2.0 maintain character and style consistency across multiple video shots?

Yes — Seedance 2.0 is designed for coherent multi-shot storytelling with consistent characters, visual style, and atmosphere across scenes, reducing common AI video drift issues.

What outputs and quality levels can I expect from Seedance 2.0 videos?

Seedance 2.0 can generate cinematic-grade videos (up to 2K resolution) with native audio, synchronized dialogue, and natural motion synthesis, typically in clips of 5–60 seconds.

How does Seedance 2.0 handle audio and lip synchronization?

The model generates audio and video jointly, offering native audio-visual sync with phoneme-level lip sync in 8+ languages for natural speech and sound effects.

Is Seedance 2.0 suitable for professional creative projects like marketing or narrative shorts?

Yes — Seedance 2.0’s multimodal control, multi-shot continuity, and high fidelity output make it suitable for marketing videos, narrative shorts, ads, and other professional applications.

How do referencing assets (images, video clips) work in Seedance 2.0 prompts?

Users can upload reference assets and then describe in natural language how each should influence motion, camera movement, or stylistic elements, giving fine-grained control over the generated content.

Does Seedance 2.0 allow editing and extension of existing videos?

Yes — the model supports video extension and targeted editing like adding scenes, replacing characters, or altering specific segments while preserving unedited portions.

What are known limitations or typical generation lengths with Seedance 2.0?

Typical output lengths range from ~5 to ~60 seconds per video, and combining many assets or high-resolution settings can increase generation time.

Kimi K3 is now live on CometAPI →

Seedance 2-0

Per Sekund:$0.056

Udgivet:Apr 19, 2026

Seedance 2.0 er ByteDance’s næste generations multimodale videogrundmodel med fokus på cinematisk, fortællende videogenerering med flere indstillinger. I modsætning til enkelt‑indstillings tekst‑til‑video‑demoer lægger Seedance 2.0 vægt på referencebaseret styring (billeder, korte klip, lyd), sammenhængende karakter- og stilkonsistens på tværs af indstillinger samt indbygget lyd-/videosynkronisering — med det mål at gøre AI‑video nyttig for professionelle kreative og arbejdsgange til previsualisering.

Populær

Kommersiel brug

Playground for Seedance 2-0

Udforsk Seedance 2-0's Playground — et interaktivt miljø til at teste modeller og køre forespørgsler i realtid. Prøv prompts, juster parametre og iterer øjeblikkeligt for at fremskynde udvikling og validere use cases.

Tekniske specifikationer for Doubao Seedance 2.0

Element	Detaljer
Produktfamilie	ByteDance Seedance-videogenereringsfamilie
CometAPI-slug	doubao-seedance-2-0
Udbyder	ByteDance / BytePlus ModelArk
Modeltype	Videogenereringsmodel
Inputmodaliteter	Tekst, billede
Outputtype	MP4-video
Outputopløsning	480p, 720p, 1080p
Sideforhold	21:9, 16:9, 4:3, 1:1, 3:4, 9:16
Outputvarighed	4–15 sekunder
API-stil	Asynkron opgavebaseret videogenerering

Hvad er Doubao Seedance 2.0?

Doubao Seedance 2.0 er ByteDance’s flagskibs-multimodale videogenereringsmodel (udgivet februar 2026). Den bruger en samlet lyd- og videogenereringsarkitektur, der muliggør præcis instruktørniveau-kontrol ved samtidig at referere til flere billeder, videoklip og lydfiler. Den udmærker sig ved at producere filmiske, bevægelsesstabile videoer med indbygget lydsynkronisering — ideel for professionelle skabere, der har brug for realistisk fysik, konsistente karakterer og kompleks scenekomposition.

Hovedfunktioner i Doubao Seedance 2.0 på CometAPI

Samlet multimodal generering: Accepterer tekst-, billede-, lyd- og video-input i én arbejdsgang, så brugere kan instruere en scene med rigere referencemateriale end en ren tekstgenerator.
Instruktørniveau-kontrol: Modellen er bygget til at følge prompts for performance, lys, skygge, kamerabevægelse, bevægelsesrytme og lydkarakteristika.
Videoredigering og udvidelse: Seedance 2.0 er ikke begrænset til førstegennemløbs-generering; den understøtter redigering af specifikke klip, karakterer, handlinger og storylines samt fremad-/bagudrettet udvidelse.
Stærk audiovisuelt sync: ByteDance fremhæver to-kanals stereo-output og synkroniseret baggrundsmusik, omgivelseslyd og voiceover-adfærd for mere medrivende resultater.
Høj styrbarhed i komplekse scener: Fremhæver forbedret håndtering af komplekse interaktioner, bevægelsesscener og fysisk plausibilitet sammenlignet med Seedance 1.5.
Produktorienteret output: API’et understøtter konfigurerbart sideforhold, varighed og vandmærkeindstillinger, hvilket gør modellen praktisk til gentagelige indholdsarbejdsgange.

Benchmark-ydelse

ByteDance oplyser, at Seedance 2.0 bruger den interne SeedVideoBench-2.0-evaluering og placerer sig i førerfeltet på tværs af tekst-til-video, billede-til-video og multimodale opgavedimensioner. Modellen forbedrer genereringskvalitet, fysisk nøjagtighed, realisme og styrbarhed sammenlignet med Seedance 1.5, især i komplekse interaktions- og bevægelsesscener.

Seedance 2.0 API og Seedance 2.0 Fast API-tilstande på CometAPI

Model	Modelnavn i CometAPI	Bedst til	Hovedforskel
Seedance 2.0	doubao-seedance-2-0	Multimodal videogenerering i højeste kvalitet	Bredeste referencestak og stærkeste styrbarhed i Seedance 2.0-familien.
Seedance 2.0 fast	doubao-seedance-2-0-fast	Hurtigere produktion, når der kan byttes lidt kvalitet for hastighed	ByteDance siger, at den bevarer de samme modelkapabiliteter som Seedance 2.0, men genererer hurtigere.
Seedance 1.5 Pro	doubao-seedance-1-5-pro	Tidligere generations lyd- og videogenerering	Seedance 1.5 Pro er positioneret som en indbygget fælles lyd- og videomodel, mens Seedance 2.0 udvider den multimodale reference- og redigeringsstak.

Styrker vs. konkurrenter:

vs Kling 3.0: Bedre multimodal referencestyring og indbygget lyd.
vs Sora 2: Overlegen referencepræcision og multi-shot-kapacitet; en smule lavere maks. varighed.
vs Veo 3.1: Stærkere i karakterkonsistens og prompt-overholdelse til reference-tunge arbejdsgange; Veo fører i rå filmisk grading og længere klip.

Udmærker sig ved menneskecentrerede og performance-drevne videoer.

Prøv Seedance 2.0 AI-videogenerator på CometAPI

Trin 1: Trin 1: Tilmeld dig for at få en API-nøgle

Start direkte i CometAPI Playground for at afprøve Doubao Seedance 2.0 uden opsætning eller kodning. Log blot ind på din CometAPI-konto, gå til modelsiden for doubao-seedance-2-0, upload dine referencebilleder, korte videoklip eller lydfiler, tilføj en beskrivende prompt, og generér forhåndsvisningsvideoer med det samme. Dette er den hurtigste måde at forstå, hvordan Seedance 2.0 håndterer bevægelseskonsistens, karakterudseende, kamerabevægelser og indbygget lydsynkronisering.

Trin 2: Få API-adgang til Seedance 2.0 på CometAPI

Opret eller brug din eksisterende CometAPI-nøgle for at få fuld adgang til Doubao Seedance 2.0. Efter login skal du navigere til afsnittet API Token i din konsol, generere en ny nøgle og kopiere den. Besøg derefter modelsiden for Seedance 2.0 og API-dokumentation på CometAPI for at gennemgå de understøttede parametre til integration i dine projekter.

Trin 3: Send din første anmodning med Seedance 2.0 API

Brug CometAPI-endpointet til at indsende din første videogenereringsanmodning med en klar tekstprompt og valgfrie referencefiler (billeder, video eller lyd). Systemet behandler opgaven asynkront, returnerer et opgave-ID og stiller en downloadbar MP4-video til rådighed, når den er færdig. Du kan derefter forfine dine prompts, justere indstillinger og skalere op til marketingvideoer, indhold til sociale medier, automatiserede videopipelines eller kreative applikationer drevet af Seedance 2.0.

Denne arbejdsgang giver dig hurtige eksperimenter i Playground og en smidig overgang til produktionsbrug via CometAPI.

Topgrunde til at vælge Seedance 2.0 på CometAPI

Hvorfor bruge CometAPI til Seedance 2.0

Direkte adgang via API eller Playground
Nem parameterstyring (varighed, opløsning, format)
Understøtter både tekst-til-video og billede-til-video arbejdsgange
Indbygget jobhåndtering til asynkron videogenerering

Ensartet og udviklervenlig API

CometAPI tilbyder et rent, standardiseret endpoint, der fungerer problemfrit med velkendte formater i OpenAI-stil eller dedikerede videotask-endpoints. Du får ligetil oprettelse af opgaver, polling og MP4-downloads uden at skulle håndtere Volcengine's komplekse godkendelse eller regionale begrænsninger.

Omkostningseffektiv prissætning

CometAPI leverer typisk mere konkurrencedygtige priser pr. sekund end direkte udbydere, så du kan generere videoer i høj kvalitet i stor skala til marketing, sociale medier eller automatiserede pipelines uden at sprænge budgettet.

Hurtig test med Playground

Eksperimentér straks i CometAPI Playground. Upload referencebilleder, videoer og lydfiler, justér prompts, og forhåndsvis resultater på få minutter — perfekt til at iterere på stil, bevægelse og lydsynkronisering, før du går i produktion.

Kort sagt, hvis du vil have den kreative kraft i Seedance 2.0 — bedst-i-klassen referencestyring, naturlig bevægelse og indbygget lyd — uden besværet ved direkte adgang til ByteDance, er CometAPI i øjeblikket en af de bedste platforme at bruge.

FAQ

Priser for Seedance 2-0

Udforsk konkurrencedygtige priser for Seedance 2-0, designet til at passe til forskellige budgetter og brugsbehov. Vores fleksible planer sikrer, at du kun betaler for det, du bruger, hvilket gør det nemt at skalere, efterhånden som dine krav vokser. Opdag hvordan Seedance 2-0 kan forbedre dine projekter, mens omkostningerne holdes håndterbare.

doubao-seedance Video Generation Pricing

Parameters

Parameter	Description
Duration (seconds)	4–15 seconds, default 5 seconds
size	Refer to documentation

*1080p is available for all models except doubao-seedance-2-0-fast

Pricing (Per Second)

Model	480p	720p	1080p	4k
`doubao-seedance-2-0`	$0.056	$0.1216	$0.2992	$0.624
`doubao-seedance-2-0-fast`	$0.0448	$0.096	—	—
`doubao-seedance-1-5-pro`	$0.0192	$0.0416	$0.0928	—
`doubao-seedance-1-0-pro`	$0.0192	$0.0416	$0.0976	—
`doubao-seedance-1-0-pro-fast`	$0.008	$0.016	$0.0384	—

💡 Billed per second. Total cost = price per second × video duration (seconds). Duration range: 4–15 seconds.

Eksempelkode og API til Seedance 2-0

Få adgang til omfattende eksempelkode og API-ressourcer for Seedance 2-0 for at strømline din integrationsproces. Vores detaljerede dokumentation giver trin-for-trin vejledning, der hjælper dig med at udnytte det fulde potentiale af Seedance 2-0 i dine projekter.

POST

/v1/videos

import json
import os
import time

import requests

# Get your CometAPI key from https://www.cometapi.com/console/token, and paste it here
COMETAPI_KEY = os.environ.get("COMETAPI_KEY") or "<YOUR_COMETAPI_KEY>"
BASE_URL = "https://api.cometapi.com"
OUTPUT_DIR = "./output"
POLL_INTERVAL_SECONDS = 10
RETRY_DELAY_SECONDS = 5
MAX_CREATE_ATTEMPTS = 5
MAX_QUERY_ATTEMPTS = 3
TERMINAL_STATUSES = {"success", "completed", "failed", "error"}
SUCCESS_STATUSES = {"success", "completed"}

def is_progress_complete(progress):
    if isinstance(progress, int):
        return progress >= 100
    if isinstance(progress, float):
        return progress >= 100
    if isinstance(progress, str):
        try:
            return float(progress.rstrip("%")) >= 100
        except ValueError:
            return False
    return False

def is_transient_status(status_code):
    return status_code == 429 or 500 <= status_code < 600

def create_task(files):
    for attempt in range(1, MAX_CREATE_ATTEMPTS + 1):
        response = requests.post(
            f"{BASE_URL}/v1/videos",
            headers=headers,
            files=files,
            timeout=30,
        )
        if response.ok:
            return response
        if not is_transient_status(response.status_code) or attempt == MAX_CREATE_ATTEMPTS:
            response.raise_for_status()
        print(f"Create request returned {response.status_code}, retrying...")
        time.sleep(RETRY_DELAY_SECONDS)

    raise SystemExit("Failed to create task.")

def get_task(task_id):
    for attempt in range(1, MAX_QUERY_ATTEMPTS + 1):
        response = requests.get(
            f"{BASE_URL}/v1/videos/{task_id}",
            headers=headers,
            timeout=15,
        )
        if response.ok:
            return response
        if not is_transient_status(response.status_code) or attempt == MAX_QUERY_ATTEMPTS:
            response.raise_for_status()
        print(f"Status request returned {response.status_code}, retrying...")
        time.sleep(RETRY_DELAY_SECONDS)

    raise SystemExit("Failed to query task.")

if COMETAPI_KEY == "<YOUR_COMETAPI_KEY>":
    print("Set COMETAPI_KEY before running this example.")
    raise SystemExit(0)

headers = {"Authorization": f"Bearer {COMETAPI_KEY}"}

create_response = create_task(
    {
        "prompt": (None, "A slow cinematic camera push across a coastal landscape at sunrise."),
        "model": (None, "doubao-seedance-2-0"),
        "seconds": (None, "5"),
        "size": (None, "16:9"),
    }
)
create_response.raise_for_status()
create_result = create_response.json()

task_id = create_result.get("id") or create_result.get("task_id")
if not task_id:
    print(json.dumps(create_result, indent=2))
    raise SystemExit("No task id returned.")

print(f"Task created: {task_id}")
print(f"Initial status: {create_result.get('status')}")

while True:
    task_response = get_task(task_id)
    task_response.raise_for_status()
    task = task_response.json()
    status = str(task.get("status") or "unknown")
    normalized_status = status.lower()
    progress = task.get("progress")
    should_try_download = normalized_status in SUCCESS_STATUSES or (
        normalized_status == "unknown" and is_progress_complete(progress)
    )

    print(f"Status: {status}, progress: {progress}")

    if should_try_download or normalized_status in TERMINAL_STATUSES:
        if should_try_download:
            video_url = task.get("video_url") or ""
            content_url = f"{BASE_URL}/v1/videos/{task_id}/content"
            output_path = os.path.join(OUTPUT_DIR, f"{task_id}.mp4")

            os.makedirs(OUTPUT_DIR, exist_ok=True)
            with requests.get(
                content_url,
                headers=headers,
                timeout=120,
                stream=True,
            ) as video_response:
                video_response.raise_for_status()
                with open(output_path, "wb") as output_file:
                    for chunk in video_response.iter_content(chunk_size=8192):
                        if chunk:
                            output_file.write(chunk)

            print(f"Video URL: {video_url}")
            print(f"Content endpoint: {content_url}")
            print(f"Saved to {output_path}")
            print(f"File size: {os.path.getsize(output_path)} bytes")
        else:
            print(json.dumps(task, indent=2))
            raise SystemExit(1)
        break

    time.sleep(POLL_INTERVAL_SECONDS)

Python Code Example

import json
import os
import time

import requests

# Get your CometAPI key from https://www.cometapi.com/console/token, and paste it here
COMETAPI_KEY = os.environ.get("COMETAPI_KEY") or "<YOUR_COMETAPI_KEY>"
BASE_URL = "https://api.cometapi.com"
OUTPUT_DIR = "./output"
POLL_INTERVAL_SECONDS = 10
RETRY_DELAY_SECONDS = 5
MAX_CREATE_ATTEMPTS = 5
MAX_QUERY_ATTEMPTS = 3
TERMINAL_STATUSES = {"success", "completed", "failed", "error"}
SUCCESS_STATUSES = {"success", "completed"}


def is_progress_complete(progress):
    if isinstance(progress, int):
        return progress >= 100
    if isinstance(progress, float):
        return progress >= 100
    if isinstance(progress, str):
        try:
            return float(progress.rstrip("%")) >= 100
        except ValueError:
            return False
    return False


def is_transient_status(status_code):
    return status_code == 429 or 500 <= status_code < 600


def create_task(files):
    for attempt in range(1, MAX_CREATE_ATTEMPTS + 1):
        response = requests.post(
            f"{BASE_URL}/v1/videos",
            headers=headers,
            files=files,
            timeout=30,
        )
        if response.ok:
            return response
        if not is_transient_status(response.status_code) or attempt == MAX_CREATE_ATTEMPTS:
            response.raise_for_status()
        print(f"Create request returned {response.status_code}, retrying...")
        time.sleep(RETRY_DELAY_SECONDS)

    raise SystemExit("Failed to create task.")


def get_task(task_id):
    for attempt in range(1, MAX_QUERY_ATTEMPTS + 1):
        response = requests.get(
            f"{BASE_URL}/v1/videos/{task_id}",
            headers=headers,
            timeout=15,
        )
        if response.ok:
            return response
        if not is_transient_status(response.status_code) or attempt == MAX_QUERY_ATTEMPTS:
            response.raise_for_status()
        print(f"Status request returned {response.status_code}, retrying...")
        time.sleep(RETRY_DELAY_SECONDS)

    raise SystemExit("Failed to query task.")

if COMETAPI_KEY == "<YOUR_COMETAPI_KEY>":
    print("Set COMETAPI_KEY before running this example.")
    raise SystemExit(0)

headers = {"Authorization": f"Bearer {COMETAPI_KEY}"}

create_response = create_task(
    {
        "prompt": (None, "A slow cinematic camera push across a coastal landscape at sunrise."),
        "model": (None, "doubao-seedance-2-0"),
        "seconds": (None, "5"),
        "size": (None, "16:9"),
    }
)
create_response.raise_for_status()
create_result = create_response.json()

task_id = create_result.get("id") or create_result.get("task_id")
if not task_id:
    print(json.dumps(create_result, indent=2))
    raise SystemExit("No task id returned.")

print(f"Task created: {task_id}")
print(f"Initial status: {create_result.get('status')}")

while True:
    task_response = get_task(task_id)
    task_response.raise_for_status()
    task = task_response.json()
    status = str(task.get("status") or "unknown")
    normalized_status = status.lower()
    progress = task.get("progress")
    should_try_download = normalized_status in SUCCESS_STATUSES or (
        normalized_status == "unknown" and is_progress_complete(progress)
    )

    print(f"Status: {status}, progress: {progress}")

    if should_try_download or normalized_status in TERMINAL_STATUSES:
        if should_try_download:
            video_url = task.get("video_url") or ""
            content_url = f"{BASE_URL}/v1/videos/{task_id}/content"
            output_path = os.path.join(OUTPUT_DIR, f"{task_id}.mp4")

            os.makedirs(OUTPUT_DIR, exist_ok=True)
            with requests.get(
                content_url,
                headers=headers,
                timeout=120,
                stream=True,
            ) as video_response:
                video_response.raise_for_status()
                with open(output_path, "wb") as output_file:
                    for chunk in video_response.iter_content(chunk_size=8192):
                        if chunk:
                            output_file.write(chunk)

            print(f"Video URL: {video_url}")
            print(f"Content endpoint: {content_url}")
            print(f"Saved to {output_path}")
            print(f"File size: {os.path.getsize(output_path)} bytes")
        else:
            print(json.dumps(task, indent=2))
            raise SystemExit(1)
        break

    time.sleep(POLL_INTERVAL_SECONDS)

JavaScript Code Example

import fs from "fs";
import path from "path";

// Get your CometAPI key from https://www.cometapi.com/console/token, and paste it here
const api_key = process.env.COMETAPI_KEY || "<YOUR_COMETAPI_KEY>";
const base_url = "https://api.cometapi.com";
const output_dir = "./output";
const poll_interval_ms = 10_000;
const retry_delay_ms = 5_000;
const max_create_attempts = 5;
const max_query_attempts = 3;
const terminal_statuses = new Set(["success", "completed", "failed", "error"]);
const success_statuses = new Set(["success", "completed"]);

function is_progress_complete(progress) {
  if (typeof progress === "number") {
    return progress >= 100;
  }

  if (typeof progress === "string") {
    const numeric = Number(progress.replace(/%$/, ""));
    return Number.isFinite(numeric) && numeric >= 100;
  }

  return false;
}

function is_transient_status(status) {
  return status === 429 || status >= 500;
}

async function fetch_with_retry(url, options, attempts, label) {
  for (let attempt = 1; attempt <= attempts; attempt += 1) {
    const response = await fetch(url, options);
    if (response.ok) {
      return response;
    }

    if (!is_transient_status(response.status) || attempt === attempts) {
      return response;
    }

    console.log(`${label} returned ${response.status}, retrying...`);
    await new Promise((resolve) => setTimeout(resolve, retry_delay_ms));
  }

  throw new Error(`${label} failed`);
}

if (api_key === "<YOUR_COMETAPI_KEY>") {
  console.log("Set COMETAPI_KEY before running this example.");
  process.exit(0);
}

const headers = {
  Authorization: `Bearer ${api_key}`,
};

const form = new FormData();
form.set("prompt", "A slow cinematic camera push across a coastal landscape at sunrise.");
form.set("model", "doubao-seedance-2-0");
form.set("seconds", "5");
form.set("size", "16:9");

const create_response = await fetch_with_retry(`${base_url}/v1/videos`, {
  method: "POST",
  headers,
  body: form,
}, max_create_attempts, "Create request");

if (!create_response.ok) {
  console.log(await create_response.text());
  process.exit(1);
}

const create_result = await create_response.json();
const task_id = create_result.id || create_result.task_id;

if (!task_id) {
  console.log(JSON.stringify(create_result, null, 2));
  process.exit(1);
}

console.log(`Task created: ${task_id}`);
console.log(`Initial status: ${create_result.status}`);

while (true) {
  const task_response = await fetch_with_retry(`${base_url}/v1/videos/${task_id}`, {
    headers,
  }, max_query_attempts, "Status request");

  if (!task_response.ok) {
    console.log(await task_response.text());
    process.exit(1);
  }

  const task = await task_response.json();
  const status = String(task.status || "unknown");
  const normalized_status = status.toLowerCase();
  const progress = task.progress;
  const should_try_download = success_statuses.has(normalized_status) || (
    normalized_status === "unknown" && is_progress_complete(progress)
  );

  console.log(`Status: ${status}, progress: ${progress}`);

  if (should_try_download || terminal_statuses.has(normalized_status)) {
    if (should_try_download) {
      const video_url = task.video_url || "";
      const content_url = `${base_url}/v1/videos/${task_id}/content`;
      const output_path = path.join(output_dir, `${task_id}.mp4`);

      if (!fs.existsSync(output_dir)) {
        fs.mkdirSync(output_dir, { recursive: true });
      }

      const video_response = await fetch(content_url, { headers });
      if (!video_response.ok) {
        console.log(await video_response.text());
        process.exit(1);
      }

      const video_buffer = Buffer.from(await video_response.arrayBuffer());
      fs.writeFileSync(output_path, video_buffer);

      console.log(`Video URL: ${video_url}`);
      console.log(`Content endpoint: ${content_url}`);
      console.log(`Saved to ${output_path}`);
      console.log(`File size: ${fs.statSync(output_path).size} bytes`);
    } else {
      console.log(JSON.stringify(task, null, 2));
      process.exit(1);
    }
    break;
  }

  await new Promise((resolve) => setTimeout(resolve, poll_interval_ms));
}

Curl Code Example

#!/bin/bash

set -euo pipefail

# Get your CometAPI key from https://www.cometapi.com/console/token
# Export it as: export COMETAPI_KEY="your-key-here"

if [[ -z "${COMETAPI_KEY:-}" ]]; then
  echo "Set COMETAPI_KEY before running this example."
  exit 0
fi

BASE_URL="https://api.cometapi.com"
OUTPUT_DIR="./output"
POLL_INTERVAL_SECONDS=10
RETRY_DELAY_SECONDS=5
MAX_CREATE_ATTEMPTS=5
MAX_QUERY_ATTEMPTS=3

is_progress_complete() {
  local progress="$1"
  local normalized="${progress%%%}"

  if [[ -z "$normalized" ]]; then
    return 1
  fi

  [[ "$normalized" =~ ^[0-9]+([.][0-9]+)?$ ]] || return 1
  awk -v value="$normalized" 'BEGIN { exit !(value >= 100) }'
}

create_task() {
  local attempt=1
  while (( attempt <= MAX_CREATE_ATTEMPTS )); do
    local response
    local status_code
    response=$(curl -sS -w $'\n%{http_code}' "${BASE_URL}/v1/videos" \
      -H "Authorization: Bearer $COMETAPI_KEY" \
      -F 'prompt="A slow cinematic camera push across a coastal landscape at sunrise."' \
      -F 'model="doubao-seedance-2-0"' \
      -F 'seconds="5"' \
      -F 'size="16:9"')

    status_code=$(echo "$response" | tail -n 1)
    CREATE_RESPONSE=$(echo "$response" | sed '$d')

    if [[ "$status_code" =~ ^2 ]]; then
      return 0
    fi

    if [[ "$status_code" == "429" || "$status_code" =~ ^5 ]] && (( attempt < MAX_CREATE_ATTEMPTS )); then
      echo "Create request returned ${status_code}, retrying..."
      sleep "$RETRY_DELAY_SECONDS"
      (( attempt += 1 ))
      continue
    fi

    echo "$CREATE_RESPONSE"
    return 1
  done
}

get_task() {
  local task_id="$1"
  local attempt=1
  while (( attempt <= MAX_QUERY_ATTEMPTS )); do
    local response
    local status_code
    response=$(curl -sS -w $'\n%{http_code}' "${BASE_URL}/v1/videos/${task_id}" \
      -H "Authorization: Bearer $COMETAPI_KEY")

    status_code=$(echo "$response" | tail -n 1)
    TASK_RESPONSE=$(echo "$response" | sed '$d')

    if [[ "$status_code" =~ ^2 ]]; then
      return 0
    fi

    if [[ "$status_code" == "429" || "$status_code" =~ ^5 ]] && (( attempt < MAX_QUERY_ATTEMPTS )); then
      echo "Status request returned ${status_code}, retrying..."
      sleep "$RETRY_DELAY_SECONDS"
      (( attempt += 1 ))
      continue
    fi

    echo "$TASK_RESPONSE"
    return 1
  done
}

create_task

TASK_ID=$(echo "$CREATE_RESPONSE" | jq -r '.id // .task_id // empty')

if [[ -z "$TASK_ID" ]]; then
  echo "$CREATE_RESPONSE" | jq .
  echo "No task id returned."
  exit 1
fi

echo "Task created: $TASK_ID"
echo "Initial status: $(echo "$CREATE_RESPONSE" | jq -r '.status // empty')"

while true; do
  get_task "$TASK_ID"

  STATUS=$(echo "$TASK_RESPONSE" | jq -r '.status // empty')
  NORMALIZED_STATUS=$(echo "$STATUS" | tr '[:upper:]' '[:lower:]')
  PROGRESS=$(echo "$TASK_RESPONSE" | jq -r '.progress // empty')
  SHOULD_TRY_DOWNLOAD=0

  if [[ "$NORMALIZED_STATUS" == "success" || "$NORMALIZED_STATUS" == "completed" ]]; then
    SHOULD_TRY_DOWNLOAD=1
  elif [[ "$NORMALIZED_STATUS" == "unknown" ]] && is_progress_complete "$PROGRESS"; then
    SHOULD_TRY_DOWNLOAD=1
  fi

  echo "Status: ${STATUS}, progress: ${PROGRESS}"

  if [[ "$SHOULD_TRY_DOWNLOAD" == "1" || "$NORMALIZED_STATUS" == "failed" || "$NORMALIZED_STATUS" == "error" ]]; then
    if [[ "$SHOULD_TRY_DOWNLOAD" == "1" ]]; then
      VIDEO_URL=$(echo "$TASK_RESPONSE" | jq -r '.video_url // empty')
      CONTENT_URL="${BASE_URL}/v1/videos/${TASK_ID}/content"
      OUTPUT_PATH="${OUTPUT_DIR}/${TASK_ID}.mp4"

      mkdir -p "$OUTPUT_DIR"
      curl -fsS "$CONTENT_URL" \
        -H "Authorization: Bearer $COMETAPI_KEY" \
        -o "$OUTPUT_PATH"

      if [[ ! -s "$OUTPUT_PATH" ]]; then
        echo "Failed to download video"
        exit 1
      fi

      echo "Video URL: ${VIDEO_URL}"
      echo "Content endpoint: ${CONTENT_URL}"
      echo "Saved to ${OUTPUT_PATH}"
      echo "File size: $(wc -c < "$OUTPUT_PATH" | tr -d ' ') bytes"
    else
      echo "$TASK_RESPONSE" | jq .
      exit 1
    fi
    break
  fi

  sleep "$POLL_INTERVAL_SECONDS"
done

Versioner af Seedance 2-0

Årsagen til, at Seedance 2-0 har flere øjebliksbilleder kan omfatte potentielle faktorer såsom variationer i output efter opdateringer, der kræver ældre øjebliksbilleder for konsistens, at give udviklere en overgangsperiode til tilpasning og migration, og at forskellige øjebliksbilleder svarer til globale eller regionale slutpunkter for at optimere brugeroplevelsen. For detaljerede forskelle mellem versioner, henvises der til den officielle dokumentation.

version
doubao-seedance-2-0-fast
doubao-seedance-2-0