The Seedance 2.0 API, Built for Production (4K is Now Supported)

The Seedance 2.0 API is ByteDance's multimodal video model, served through Atlas Cloud. From a single endpoint you can generate 4–15 second clips at 480p or 720p with synchronized native audio. It runs three modes: text-to-video (prompt only), image-to-video (animate a still, with optional first- and last-frame control), and reference-to-video (feed it images, video, or audio to lock character identity, visual style, or camera motion). It's best for short cinematic shots, product/fashion showcases, and character-consistent scenes — not for long-form (15s+) single generations.

1. Create a free Atlas Cloud account to use the Seedance 2.0 API.

2. Open Dashboard → API Keys and copy your key.

3. Paste it into the quickstart above and run it. There's no waitlist or manual approval, so step 1 to a finished video is usually under five minutes. Set the key as an environment variable (ATLAS_API_KEY) rather than hardcoding it.

Use Seedance 2.0 Fast ($0.09/s) for iteration, drafts, social/UGC content, and high-volume batch jobs where speed and cost matter most. Use Seedance 2.0 Standard ($0.112/s) for final/hero shots where you want maximum fidelity and motion stability. A common workflow: prototype prompts on Fast, then re-render the winner on Standard.

Images: JPEG, PNG, WebP Audio: WAV, MP3 Reference video: MP4, MOV Text: prompt string Output is always MP4 with an audio track. Supported aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 — set this at request time to match your target platform (e.g. 9:16 for Reels/TikTok, 16:9 for YouTube).

Pricing is per second of output video, the same across all resolutions and modes: $0.112/s standard and $0.09/s Fast. Multiply by clip length:

5s clip → $0.45 (Fast) / $0.56 (standard)

10s clip → $0.90 (Fast) / $1.12 (standard)

There's no subscription or monthly minimum — you pay only for the seconds you generate, and failed generations aren't charged.

Yes. Atlas Cloud runs the Seedance 2.0 API on infrastructure built for production, not just prototyping: a 99.99% uptime guarantee, SOC 2 certification, HIPAA-compliant infrastructure, RBAC, and compliance-ready logging. Rate limits vary by account tier and model; if you hit a 429 Too Many Requests error, contact support to raise your limits. For dedicated capacity, private-cloud deployment, and custom SLAs, see the Enterprise plan.

Yes. The Seedance 2.0 API generates lifelike human characters with strong identity and style consistency across shots, and can use images, audio, and video as references to guide them. The important limit: it can render realistic generated/virtual humans, but you cannot make it depict a specific real person by uploading their photo (see below). For consistent characters, use the platform's virtual-portrait/reference workflow rather than a real face.

No. Per ByteDance's official platforms, the Seedance 2.0 series does not support directly uploading reference images or videos that contain real human faces (Volcengine), and "generating videos using real human images is not supported" (BytePlus). The model runs face detection on reference uploads and rejects photorealistic real faces.

To prevent unauthorized use of a real person's likeness and to limit deepfake/legal risk. ByteDance added the real-face upload restriction as part of Seedance 2.0's global relaunch, following likeness/"virtual piracy" disputes involving Hollywood studios; it also suspended a related face-to-voice feature in February 2026 over privacy risk.

The officially supported route is to use the platform-provided virtual portrait / character library, where each virtual portrait carries an asset ID you pass via URI in the API — not uploading your own face image. ByteDance's docs recommend "use virtual characters provided by the platform." Uploading your own AI-generated or stylized portrait is not officially documented as supported; some third-party testing reports that fully synthetic/illustrated faces pass the filter, but ByteDance does not guarantee this.