Seedance vs Seedream: ByteDance's video and image models
Seedance is ByteDance Seed's video model family; Seedream is its image model family. What each makes, takes, and costs through one API.

Seedance makes video and Seedream makes images. Both are model families from ByteDance Seed: Seedance turns a prompt, images, audio or video into a video clip, and Seedream generates and edits still images.
ByteDance facts below come from its own Seedance 2.0 and Seedream 4.5 pages. The API facts come from Sume, which serves both families through one key: the Video generation and Image API docs and its pricing code, all read 2026-09-28.
What is Seedance?
ByteDance's model page says Seedance 2.0 “adopts a unified multimodal audio-video joint generation architecture that supports text, image, audio, and video inputs”: it makes a clip with its sound, guided by the media you give it. The same Seed site lists Seedance 2.5 among its models.
On Sume, Seedance is seedance-2.5, seedance-2-mini, seedance-2, seedance-2-fast on POST /v1/videos. seedance-2.5 makes 4–30 second clips and the 2.0 ids stop at 15 seconds; Seedance 2.0 Fast vs Seedance 2.0 compares the 2.0 ids.
What is Seedream?
Seedream is the image family. ByteDance's page for Seedream 4.5 describes multi-image editing that “strictly preserves the details of the reference images” and improved “typography and dense text rendering”. It returns pictures, not motion.
On Sume, Seedream is bytedance-seed/seedream-4.5, bytedance-seed/seedream-5-lite, bytedance-seed/seedream-4 on POST /v1/images. Seedream 4.5 API on Sume covers edits, reference images and sizes.
How do Seedance and Seedream differ in an API?
They sit on different endpoints, take different inputs, return results differently, and bill in different units:
| Seedance | Seedream | |
|---|---|---|
| What it makes | Video clips, with optional sound | Still images |
| Sume endpoint | POST /v1/videos | POST /v1/images |
| Model ids on Sume | seedance-2.5, seedance-2-mini, seedance-2, seedance-2-fast | bytedance-seed/seedream-4.5, bytedance-seed/seedream-5-lite, bytedance-seed/seedream-4 |
| Inputs | A prompt, first and last frames, and image, video or audio references | A prompt and optional reference images in input_references |
| How you get the result | Asynchronous: poll the job until completed, then download | sync by default, with a wait of up to 30 seconds; longer runs return a job |
| Billed per | 1,000 video tokens ($0.00875 to $0.02675) | Completed image |
Can I use Seedream and Seedance together?
Yes, as two calls. Generate a still with Seedream, then pass it to a Seedance id as the first frame in frame_images (image-to-video) or as an image in input_references (reference-to-video). If you send both fields, frame_images takes precedence.
- Seedream result URLs on Sume are signed; copy the files you want to keep.
- Seedance frame and reference images must be reachable over public HTTPS.
- Seedream 4.5 returns one image per call when you send a reference image, whatever
nasks for (current behavior).
How is each one billed?
Seedance bills video tokens, which grow with the clip's pixels and length, at the provider's list price × 1.25, reserved from your balance when you submit. Seedream bills per image: a completed generation is billed in full, and a failed or canceled one is not billed. Each image model's GET /v1/images/models/{model_id}/endpoints record carries its pricing line, with Sume's margin already applied. Both add a 5.5% agent fee by default, and the response's usage.cost is the billed amount.
Sources
Related posts
More in Models
- Seedance vs Sora 2: specs side by side after the shutdown
Sora 2 shut down on September 24, 2026. Its clip lengths, frame sizes, inputs, extensions and edits beside Seedance 2.0 and 2.5, which you can still call.
- Sora MCP server: why it stopped working and what to use
A Sora MCP server can't make video now: OpenAI shut the Sora API down on September 24, 2026. Point your MCP client at other video models instead.
- Text to speech in Japanese: set the language to ja
For Japanese text to speech, send the script in kana and kanji and set the language to ja. Left out, Sume can read a kanji-only line as English.
- Text to speech pronunciation: how to fix wrong words
Fix text to speech pronunciation: set the text's language, use a voice recorded in it, and respell tricky words. What Sume's TTS API checks for you.
Written by Sume