ByteDance/bytedance-seedance-2-5-reference-to-video/Released Jul 2026

Seedance 2.5 - Reference to Video

Direct with your references: up to 30 images, 10 videos, and 10 audio clips guiding one generation. Real faces, real action, and native synced audio -- single takes up to 30 seconds.

FeaturedCommercial use
Video

About Seedance 2.5 - Reference to Video

ByteDance Seedance 2.5 generates video from a text prompt guided by reference images, videos, and/or audio, with synchronized audio on by default. Accepts up to 30 reference images, 10 reference videos, and 10 reference audio clips, including audio-only input. Supports generating new videos, editing videos, extending videos, and coherent outputs up to 30 seconds long at 480p, 720p, 1080p, or 4k (1080p and 4k use 10-bit color and H.265/HEVC encoding).

Draft mode renders a quick 480p preview for checking the scene, motion, and prompt. Its draft task ID then renders the final video at 1080p or 4k with the same prompt, media, and settings, and stays valid for seven days after the draft is created.

See what this model can do

A handful of real prompts and their outputs.

Cinematic
Salt Lake at Sunrise
"Wide tracking shot, mirror reflections, golden mist."
One-take
One-Take Kitchen Rush
"20-second unbroken handheld shot, native synced audio."
Realistic characters
The Potter's Hands
"Photorealistic skin, natural light, subtle performance."
Commercial
Perfume Splash
"Slow-motion product drop, crown splash, studio lighting."
UGC
Night Market Vlog
"Handheld selfie energy, lip-synced speech, natural audio."
3D Animation
Robot in the Koi Pond
"Stylized 3D render, expressive character animation, rain ambience."
Color
Holi Powder Dance
"Slow-motion powder bursts, deep blacks, rich saturated color."
Macro
Neon Chameleon
"Macro detail, shifting neon color, rain-slick city bokeh."

Ready to build with Seedance 2.5 - Reference to Video?

Try Seedance 2.5 - Reference to Video in the Workbench to prompt it, compare outputs, and iterate on prompts without writing any code. When you're ready to ship, call the same model from our API and build your own apps on top of it.

curl -sSf -X POST https://hub.oxen.ai/api/ai/videos/generate \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $OXEN_API_KEY" \
    -d '{
  "model": "bytedance-seedance-2-5-reference-to-video",
  "prompt": "A lone ox walks down an empty desert highway at golden hour, dust drifting behind it, slow cinematic camera move, warm low sunlight, shallow depth of field.",
  "omni_reference_task_type": "auto",
  "generate_audio": true,
  "aspect_ratio": "adaptive",
  "duration": "auto",
  "resolution": "720p",
  "draft": false,
  "output_format": "mp4",
  "watermark": false
}'

Request parameters

Every field Seedance 2.5 - Reference to Video accepts in the request body. See the API reference for response formats.

POSThttps://hub.oxen.ai/api/ai/videos/generate
FieldTypeRequiredDefaultDescription
promptstringYes
—
The text prompt used to generate the video. Use @Image1, @Image2, @Video1, @Audio1, etc. to reference inputs. Face-containing media is indexed before non-face media within its type.
input_face_imagesarray<string>No
—
Place reference images here when they show faces, people, or characters. Content filters may block the request otherwise.
input_imagesarray<string>No
—
Reference images that guide the generated video. Up to 30 total reference images may be used.
input_face_videosarray<string>No
—
Place reference videos here when they show faces, people, or characters. Content filters may block the request otherwise.
input_videosarray<string>No
—
Reference videos that guide the generated video. Up to 10 total reference videos may be used, each 2-30 seconds long (4-30 seconds for Edit video) with a combined duration of at most 30 seconds.
input_audiosarray<string>No
—
Reference audios that guide the generated video. Audio-only input is supported. Up to 10 reference audio clips may be used, each 2-30 seconds long with a combined duration of at most 30 seconds.
omni_reference_task_typestringNo"auto"Whether the reference video is edited or extended, or 'Auto' to let the prompt decide. 'Edit' and 'Extend' need a reference video and match the video's aspect ratio.
autoeditextend
generate_audiobooleanNotrueWhether to generate synchronized audio.
aspect_ratiostringNo"adaptive"
adaptive21:916:94:31:13:49:16
durationinteger | stringNo"auto"
Range: 4 – 30
resolutionstringNo"720p"
480p720p1080p4k
draftbooleanNofalseGenerate a quick 480p preview to check the scene and motion before the full-quality render.
draft_task_idstringNo
—
Render the final video from a draft, reusing its prompt, media, and settings. Set resolution to 1080p (the default) or 4k. Draft task IDs expire seven days after the draft is created.
output_formatstringNo"mp4"Output video format. mp4 for broad compatibility, mov for high color precision in professional post-production workflows. For video editing and extension, mov input and output is recommended.
mp4mov
watermarkbooleanNofalseWhether to add an 'AI generated' watermark to the output.