ByteDance Seedance 2.0 Mini API on E2X: 4–15s Video & Multimodal References from $0.045/s
ByteDance's Seedance 2.0 Mini (dreamina-seedance-2-0-mini-260615) is now live on the E2X API across all four video generation capabilities:
bytedance/seedance-2.0-mini/text-to-video: Generate 4 to 15 second clips from a text prompt across six aspect ratios (21:9,16:9,4:3,1:1,3:4,9:16) with native synchronized audio.bytedance/seedance-2.0-mini/image-to-video: Animate a starting image (image_url) into a 4 to 15 second clip with camera and motion control.bytedance/seedance-2.0-mini/start-end-frame-to-video: Interpolate a continuous 4 to 15 second shot between an opening frame (image_url) and a closing frame (end_frame_url).bytedance/seedance-2.0-mini/reference-to-video: Combine up to 9 reference images (@Image1..@Image9), 3 reference video clips (@Video1..@Video3), and 3 reference audio tracks (@Audio1..@Audio3) in a single request.
With our standing 25% discount off the $0.06/s list base, all four endpoints bill at the exact same per-second rate: $0.045 per second at 480p and $0.09 per second at 720p, with native audio included. A default 5-second 720p clip costs $0.45, and a 5-second 480p draft costs $0.225.
How the $0.045/s and $0.09/s pricing works
Every Seedance 2.0 Mini capability on E2X shares a list base rate of $0.06 per second (60,000 micro-cents). Our standing 25% global discount lowers that base to $0.045 per second (45,000 micro-cents). Two schema parameters multiply the discounted base:
resolution:"480p"applies a ×1 multiplier ($0.045/s, or45,000micro-cents/s), and"720p"(the default) applies a ×2 multiplier ($0.09/s, or90,000micro-cents/s).duration: any whole-second string from"4"to"15"(default"5") multiplies the per-second rate by the requested duration.
Unlike token-billed APIs where adding reference videos or enabling audio changes the multiplier, E2X bills strictly by output duration and output resolution:
has_sound(trueby default) carries no multiplier. Synchronized dialogue, ambient sound, and foley are included in the$0.045/sand$0.09/srates.- Multimodal reference inputs (
reference_urls,reference_video_urls,reference_audio_urls) add zero input-token surcharges.
Here is how E2X pricing compares against fal.ai, WaveSpeed, Replicate, and Eachlabs, checked on November 5, 2026:
| Provider | 480p / sec | 720p / sec | 5s at 720p (T2V / I2V) | 5s at 720p (with 5s video ref) | E2X savings at 720p |
|---|---|---|---|---|---|
| E2X (25% off) | $0.045 | $0.090 | $0.45 | $0.45 | Baseline |
| fal.ai | ~$0.072 | ~$0.155 | ~$0.77 | ~$0.93 | 42% lower (T2V) / 52% lower (video ref) |
| WaveSpeed | $0.060 | $0.120 | $0.60 | $0.75 | 25% lower (T2V) / 40% lower (video ref) |
| Replicate | $0.040 | $0.090 | $0.45 | $0.55 | Parity (T2V) / 18% lower (video ref) |
| Eachlabs | $0.035–$0.039 | $0.075–$0.087 | $0.38–$0.44 | Varies | Unified dual-provider chain on E2X |
Rate comparison checked November 5, 2026. Provider rates and terms can change; see our pricing page for live figures.
Full parameter schema across all four capabilities
All four Seedance 2.0 Mini routes accept jobs through POST https://api.e2x.ai/v1/jobs/submit. Pass the capability slug in model and the parameters below in input:
| Parameter | Endpoints | Type | Required | Default | Supported values |
|---|---|---|---|---|---|
prompt | All 4 endpoints | string | Yes | None | Up to 2,000 characters. Supports @Image1..@Image9, @Video1..@Video3, @Audio1..@Audio3 on reference-to-video. |
image_url | image-to-video, start-end-frame-to-video | string (URI) | Yes | None | Publicly reachable HTTPS URL for the opening frame. |
end_frame_url | start-end-frame-to-video | string (URI) | Yes | None | Publicly reachable HTTPS URL for the closing frame. |
reference_urls | reference-to-video | string[] | At least 1 image or video | None | Up to 9 image URLs (@Image1..@Image9). |
reference_video_urls | reference-to-video | string[] | Optional | None | Up to 3 video URLs (@Video1..@Video3). |
reference_audio_urls | reference-to-video | string[] | Optional | None | Up to 3 audio URLs (@Audio1..@Audio3). Requires at least one image or video reference. |
aspect_ratio | text-to-video, reference-to-video | string | No | "16:9" | "21:9", "16:9", "4:3", "1:1", "3:4", "9:16". |
resolution | All 4 endpoints | string | No | "720p" | "480p" ($0.045/s), "720p" ($0.09/s). |
duration | All 4 endpoints | string | No | "5" | "4", "5", "6", "7", "8", "9", "10", "11", "12", "13", "14", "15". |
has_sound | All 4 endpoints | boolean | No | true | true (synchronized native audio) or false (silent MP4). |
seed | All 4 endpoints | integer | No | Random | Non-negative integer (0 or higher) for reproducible motion. |
Dual-provider architecture and portrait moderation bypass
When you call bytedance/seedance-2.0-mini/* on E2X, your job executes against a two-stage provider chain configured in our worker:
- Primary provider (BytePlus ModelArk direct API): Connects directly to ByteDance's
dreamina-seedance-2-0-mini-260615model onark.ap-southeast.bytepluses.com. - Automatic privacy-asset registration for real faces: BytePlus ModelArk's default moderation rejects input images or reference videos containing real human faces with
InputImageSensitiveContentDetected.PrivacyInformation. When our worker detects this error code, it automatically registers the visual inputs into BytePlus's private AIGC Asset Library withModeration.Strategy = "Skip", rewrites the media URLs toasset://<id>, resubmits the task, and deletes the temporary assets in afinallyblock as soon as polling completes. - Secondary fallback (Eachlabs): If the primary upstream times out or returns a service error, the worker transparently fails over to Eachlabs (
bytedance-seedance-2-0-mini-*) without requiring a retry from your application.
Code examples: Text-to-video and multimodal reference-to-video
1. Submitting a 10-second ultrawide text-to-video job (curl)
Here is a curl request calling bytedance/seedance-2.0-mini/text-to-video in 21:9 aspect ratio at 480p draft resolution (10s × $0.045 = $0.45):
curl -X POST https://api.e2x.ai/v1/jobs/submit \
-H "Authorization: Bearer $E2X_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "bytedance/seedance-2.0-mini/text-to-video",
"input": {
"prompt": "Low-angle tracking shot ahead of a classic silver coupe driving through an alpine pass at sunrise. Golden light glints off wet asphalt and snowbanks. Deep exhaust note echoing off rock walls and wind rushing past.",
"aspect_ratio": "21:9",
"resolution": "480p",
"duration": "10",
"has_sound": true,
"seed": 77410
}
}'
2. Combining @Image1, @Video1, and @Audio1 in TypeScript
On bytedance/seedance-2.0-mini/reference-to-video, you can lock character identity from images, camera choreography from a reference video, and vocal cadence from an audio clip in one call:
const E2X_API_KEY = process.env.E2X_API_KEY!;
const headers = {
Authorization: `Bearer ${E2X_API_KEY}`,
"Content-Type": "application/json",
};
async function runMultimodalReference() {
const submitRes = await fetch("https://api.e2x.ai/v1/jobs/submit", {
method: "POST",
headers,
body: JSON.stringify({
model: "bytedance/seedance-2.0-mini/reference-to-video",
input: {
prompt:
"The presenter in @Image1 holds the matte titanium flask from @Image2 in the sunlit studio from @Image3, speaking with the vocal rhythm of @Audio1 while the camera performs the slow arc from @Video1.",
reference_urls: [
"https://cdn.example.com/assets/presenter-front.jpg",
"https://cdn.example.com/assets/titanium-flask.jpg",
"https://cdn.example.com/assets/sunlit-studio.jpg",
],
reference_video_urls: [
"https://cdn.example.com/assets/camera-arc-ref.mp4",
],
reference_audio_urls: [
"https://cdn.example.com/assets/voice-rhythm.mp3",
],
aspect_ratio: "16:9",
resolution: "720p",
duration: "6",
has_sound: true,
},
}),
});
const { data: submitted } = await submitRes.json();
for (let i = 0; i < 240; i++) {
await new Promise((r) => setTimeout(r, 3000));
const statusRes = await fetch(
`https://api.e2x.ai/v1/jobs/${submitted.jobId}`,
{ headers }
);
const { data: job } = await statusRes.json();
if (job.status === "completed") {
return job.outputs[0].url;
}
if (job.status === "failed" || job.status === "cancelled") {
throw new Error(job.error?.message || `Job ended with ${job.status}`);
}
}
}
When to pick Seedance 2.0 Mini vs Gemini Omni 1.1 Flash vs Veo 3.1 Fast
E2X offers three video model families across the text-to-video, image-to-video, start-end-frame-to-video, and reference-to-video categories:
| Feature | Seedance 2.0 Mini | Gemini Omni 1.1 Flash | Veo 3.1 Fast |
|---|---|---|---|
| 720p rate on E2X | $0.09/s | $0.09/s | $0.01/s (silent) / $0.015/s (sound) |
| Draft tier rate | $0.045/s (480p) | $0.0297/s (360p) | None |
| Start-end-frame rate | $0.09/s (720p) / $0.045/s (480p) | $0.09/s (720p) | $0.10/s |
| Max resolution | 720p | 4K ($0.18/s) | 4K |
| Duration options | 4 to 15 seconds (1s steps) | 4, 6, 8, 10 seconds | 4, 6, 8 seconds |
| Aspect ratios | 6 (21:9, 16:9, 4:3, 1:1, 3:4, 9:16) | 2 (16:9, 9:16) | 2 (16:9, 9:16) |
| Reference inputs | 9 images + 3 videos + 3 audios | Up to 7 images | Up to 3 images |
Use Seedance 2.0 Mini whenever your workflow needs:
- Single-take durations between 11 and 15 seconds, or odd-second timings (
5s,7s,9s,12s) to fit an edit timeline. - Non-standard aspect ratios (
21:9cinema,1:1square,4:3, or3:4) without post-crop loss. - Multimodal
@Videomotion references or@Audiorhythm references alongside up to 9@Imagereferences. - Keyframe transitions (
start-end-frame-to-video) at$0.045/s(480p) or$0.09/s(720p), undercutting Veo 3.1 Fast's$0.10/sstart-end-frame rate.
Explore all four endpoints in the E2X model catalog or check the machine-readable specs for text-to-video, image-to-video, start-end-frame-to-video, and reference-to-video.
Frequently asked questions
What is the price of ByteDance Seedance 2.0 Mini on E2X?
With our standing 25% discount off the $0.06/s list rate, Seedance 2.0 Mini costs $0.045 per second at 480p and $0.09 per second at 720p across all four capabilities. A 5-second 720p clip costs $0.45, and a 5-second 480p clip costs $0.225.
How much cheaper is E2X than fal.ai and WaveSpeed for Seedance 2.0 Mini?
At 720p, E2X ($0.09/s) is 42% cheaper than fal.ai (~$0.1547/s) and 25% cheaper than WaveSpeed ($0.12/s). When you attach reference videos on reference-to-video, the savings grow to 40%–52% because E2X does not bill you for input video seconds.
Can I pass reference audio by itself on reference-to-video?
No. BytePlus ModelArk requires at least one visual reference (reference_urls or reference_video_urls) in every reference-to-video job. You can attach up to 3 audio files (reference_audio_urls) alongside at least one image or video reference.
How do I reference specific images, videos, and audio files in my prompt?
Use 1-based tags matching the order of each array: @Image1 through @Image9 for reference_urls, @Video1 through @Video3 for reference_video_urls, and @Audio1 through @Audio3 for reference_audio_urls.