Seedance 2.5 — Text, Photo, Video & Audio to Video

What is Seedance 2.5 — Text, Photo, Video & Audio to Video

Seedance 2.5 is ByteDance's latest video generation model, available on reAPI as a unified async endpoint. It turns text, images, video, and audio into short video clips of 4–30 seconds, with synced audio (speech, sound effects, music). It supports up to 30 image references, 10 video references, and 10 audio references per request, and offers five generation modes: text-to-video, image-to-video, first/last-frame, multimodal reference, and prompt-driven video editing. Output resolutions include 480p, 720p, and 1080p, with mp4 or mov formats.

How to use Seedance 2.5 — Text, Photo, Video & Audio to Video

1. Access the API: Use the reAPI platform to call the Seedance 2.5 model. The model ID is doubao-seedance-2.5-face.
2. Choose a generation mode: Select from text-to-video, image-to-video, first/last-frame, reference-to-video, or video editing. The mode is determined by the fields you send in the request.
3. Provide inputs:
   • For text-to-video: Provide a prompt (3–20000 characters).
   • For image-to-video: Pass image URLs in image_urls (first image becomes the first frame).
   • For reference-to-video: Include up to 30 images, 10 videos, and 10 audio URLs.
   • For audio generation: Set generate_audio to true (default is on).
4. Set parameters: Configure duration (4–30 seconds, or set to -1 for auto), aspect ratio (adaptive or specific), resolution (480p, 720p, 1080p), output format (mp4 or mov), watermark (off by default), and seed for reproducibility.
5. Submit the task: Use the async API to submit a generation task. You will receive a task ID.
6. Poll for results: Poll the task status every 1–2 seconds until completion. The output will include a video URL with audio.
7. Chain clips (optional): For longer videos, enable return_last_frame to get a last_frame_url and feed it into the next call to continue seamlessly.

Features of Seedance 2.5 — Text, Photo, Video & Audio to Video

• Multi-modal input: Accepts text, images, video, and audio references.
• Long video generation: Produces up to 30 seconds in a single pass (double the length of Seedance 2.0).
• Synced audio: Generates audio track with speech, sound effects, and music; supports up to 10 reference audio tracks.
• Real-person references: Supports real-person reference images and videos (subject to automated moderation).
• Auto duration: Set duration to -1 and the model picks the optimal length.
• Multiple resolutions: 480p, 720p (default), and 1080p.
• Output formats: mp4 (default) or mov for professional color.
• Prompt-driven video editing: Edit existing videos based on text prompts.
• Web search tool: Can integrate web search for enhanced generation.
• Content filter: Built-in moderation (default on) to ensure safe content.
• Per-second billing: Cost scales with output duration, resolution, and reference mode.

Use Cases of Seedance 2.5 — Text, Photo, Video & Audio to Video

• Advertising: Create short promotional clips for products or services.
• User-generated content (UGC): Generate engaging videos for social media.
• Talking-head clips: Produce videos with a person speaking, using real-person references.
• Short drama: Generate narrative video segments for storytelling.
• Video editing: Modify existing videos with text prompts (e.g., change style, add elements).
• Animate still images: Bring photos to life by providing an image and a prompt.
• Audio-to-video: Use audio tracks as reference to generate matching visuals.

Pricing

Seedance 2.5 uses a credit-based system where 1 credit equals $0.001. The cost is calculated per second of output, and varies based on:

• Output duration (4–30 seconds)
• Resolution (480p, 720p, 1080p)
• Reference mode (e.g., text-only vs. multimodal)

Pricing ranges from approximately $0.071 per second (lowest) to $0.462 per second (highest). For auto duration (duration -1), the system reserves at the 30-second cap and refunds the difference based on actual output length.

FAQ

What is Seedance 2.5? Seedance 2.5 is the next-generation video model in ByteDance's Seedance family. It turns text, images, video, and audio into short video clips. It is live on reAPI and can be used via the playground or async API.

Does Seedance 2.5 support real-person reference images? Yes. Seedance 2.5 accepts real-person reference images and videos. Reference material passes an automated review before generation; anything rejected by moderation fails with a clear error and a full refund.

How will Seedance 2.5 pricing work? Pricing is credit-based (1 credit = $0.001). The cost scales with output duration × resolution × reference mode. Auto duration (duration -1) reserves at the 30-second cap and settles to the actual output length, refunding the difference.

Which Seedance 2.5 model id do I call? One id covers everything: doubao-seedance-2.5-face. Text-to-video, image-to-video, first/last frame, and multimodal reference are selected implicitly by which fields you send.

Can Seedance 2.5 animate a still image? Yes. Pass image URLs in image_urls and Seedance 2.5 switches to image-to-video. The first image becomes the first frame; the prompt drives the motion. Up to 30 reference images per request.

Does Seedance 2.5 generate audio? Yes. Set generate_audio to true and Seedance 2.5 synthesizes an audio track that plays with the video. You can also pass up to 10 reference audio tracks in audio_urls.

How long can a Seedance 2.5 video be? Seedance 2.5 generates 4–30 seconds per request, or set duration to -1 and the model picks the best length. For longer outputs, set return_last_frame and feed the URL into the next call to chain clips without prompt drift.

Where is the full Seedance 2.5 reference? The full parameter list, task-type rules (video edits require duration -1), error codes, and runnable cURL examples live at /docs/seedance-2-5.

How do I choose between Seedance 2.5 and Seedance 2.0? Pick Seedance 2.5 for longer clips (up to 30 s vs 15 s), bigger reference sets (30 images + 10 videos + 10 audios vs 9/3/3), audio-only reference, and prompt-driven video editing. Pick Seedance 2.0 when you need 4K output or the cheaper Fast and Mini tiers — 2.5 now serves up to 1080p.