Seedance & Seedream (BytePlus ModelArk)
Official ByteDance creative models through BytePlus ModelArk:
| Model | Service | Variants |
|---|---|---|
| Seedance 2.5 | video-generate | seedance-2-5 (default); also Seedance 2.0 Standard / Fast / Mini |
| Seedream 5.0 Pro | generate / edit | seedream-5-0-pro (default); also Lite / 4.x |
Product pages: Seedance 2.5 · Seedream 5.0 Pro on ModelArk.
Environment
export ARK_API_KEY=...
# optional:
# export BYTEPLUS_ARK_BASE_URL=https://ark.ap-southeast.bytepluses.com/api/v3
Variants
Seedance (video)
--model-version | Notes |
|---|---|
seedance-2-5 / dreamina-seedance-2-5-260628 | Up to ~30s storytelling (official ModelArk id) |
seedance-2-0 / dreamina-seedance-2-0-260128 | Production 2.0 Standard |
seedance-2-0-fast / dreamina-seedance-2-0-fast-260128 | Faster 2.0 |
seedance-2-0-mini / dreamina-seedance-2-0-mini-260615 | Lightweight 2.0 |
Seedream (image)
--model-version | ModelArk model id | Notes |
|---|---|---|
seedream-5-0-pro (default) | dola-seedream-5-0-pro-260628 | Precision T2I / edit / multi-ref; 1K, 1.5K or 2K |
seedream-5-0-flash | dola-seedream-5-0-flash-260915 | Faster, cheaper 5.0 tier with the same capabilities; 1K, 1.5K or 2K |
seedream-5-0-lite | seedream-5-0-lite-260128 | Earlier 5.0 tier |
seedream-4-5 / seedream-4-0 | seedream-4-5-251128 / seedream-4-0-250828 | Prior generation |
Seedream 5.5 launched on 9 Nov 2026 in ByteDance's apps, but BytePlus ModelArk has not published an API model id for it yet. It will be added once one is listed.
CLI
# Seedance 2.5 text-to-video
opentryon video-generate --model seedance \
--prompt "A 10-second runway walk in soft studio light" \
--duration 10 --resolution 1080p --ratio 16:9
# Image-to-video
opentryon video-generate --model seedance \
--image person.jpg --prompt "Slow cinematic push-in" --duration 5
# Multimodal reference-to-video (images / videos / audio; not combinable with --image / --end-image)
opentryon video-generate --model seedance \
--prompt "The model from image 1 repeats the walk from video 1" \
--reference-image person.jpg --reference-video walk.mp4 --reference-audio score.mp3 \
--duration 12
# Seedream 5.0 Pro text-to-image
opentryon generate --model seedream \
--prompt "Editorial product shot of matte black sneakers" --size 2K
# Multi-reference edit / fusion
opentryon edit --model seedream \
--images person.jpg garment.jpg \
--prompt "Dress the person in the garment, catalogue lighting"
Python
from tryon.api.byteplus import SeedanceAdapter, SeedreamAdapter
video = SeedanceAdapter().generate_text_to_video(
prompt="Fashion model walking through a loft apartment",
duration=8,
resolution="1080p",
)
open("out.mp4", "wb").write(video)
images = SeedreamAdapter().generate_text_to_image(
prompt="Minimalist lookbook still, soft daylight",
size="2K",
)
images[0].save("out.png")
Notes
- Seedance 2.5 public ModelArk id is
dreamina-seedance-2-5-260628(CLI--model-version seedance-2-5maps to it). Activate the model in the BytePlus console before calling. - Reference-to-video sends
reference_image/reference_video/reference_audiocontent roles; reference audio needs at least one reference image or video. Local video/audio files are sent as base64 data URLs; prefer URLs for large files. - Seedance is cloud-only; there are no open weights.
- Auth uses
Authorization: Bearer $ARK_API_KEY. - Async video tasks are polled until completion, then the MP4 is downloaded.