Video GenerationActive
Seedance 2.5 Flash Reference Video
ByteDance Seedance 2.5 Flash Reference generates, edits, or extends short videos using multimodal references like images, video clips, and audio cues.
Text to VideoImage to Video
Model ID
seedance v2.5 flash reference
Provider
bytedance
Updated
Nov 16, 2026, 11:40
wiro playground—bytedance/seedance v2.5 flash reference
Updated Nov 16, 2026, 11:40
Overview
Seedance v2.5 Flash Reference is a ByteDance video model for reference-guided generation. You provide a prompt plus optional reference images, videos, and audio. The model uses those references to lock subject identity, style, motion, and timing across a clip. It outputs a single video clip with optional synchronized audio, which helps you iterate fast on shots.
What you can build
- Character-consistent shorts using a small set of look references
- Product videos that keep branding, colors, and materials consistent
- Motion-matched shots by borrowing movement from a reference clip
- Beat-synced visuals driven by a reference music or rhythm track
- Video restyles that keep the original camera and blocking
- Shot extensions that continue an existing clip past its last frame
Inputs
- A natural-language video brief that describes subject, action, scene, camera, and sound.
- Optional reference images (1–10) to guide character, wardrobe, props, or style. Use JPG, PNG, WebP, BMP, TIFF, GIF, HEIC, or HEIF. Keep each file under 30 MB.
- Optional reference video clips (1–5) to guide motion, pacing, and shot structure. Use MP4 or MOV. Keep each clip about 2–30 seconds, with a combined total around 30 seconds. Keep each file under 200 MB.
- Optional reference audio clips (1–5) to guide rhythm, timing, and voice feel. Use MP3 or WAV. Keep each clip about 2–30 seconds, with a combined total around 30 seconds. Keep each file under 15 MB.
- A way to reference your attached assets inside the prompt by index, such as Image 1, Video 1, or Audio 1.
- A target output resolution (480p or 720p).
- A target output length (either a fixed number of seconds from 4 to 30, or automatic length selection).
- An option to generate synchronized audio with the video, including ambience, effects, and speech.
- An optional aspect ratio choice. Automatic ratio selection is supported, along with common ratios like 16:9 and 9:16.
- An optional output container choice (MP4 or MOV). MP4 is the safe playback choice. MOV is better for grading and compositing.
- An optional watermark toggle to add an “AI Generated” style mark on the output.
- An optional prompt enhancement toggle that rewrites your prompt into a safer, brandless version before generation.
Outputs
- One rendered video file in MP4 or MOV.
- If audio generation is enabled, the file includes a synced soundtrack track.
- File metadata is typically returned with the result, such as a download URL, MIME type, file name, and file size.
- A generation seed value is usually returned for repeat testing. Exact repeatability is not guaranteed.
Recommended settings
- Use 480p when you need quick drafts and many iterations.
- Use 720p when the shot is close to final and details matter.
- If you’re editing or extending a reference video, keep duration on automatic so the timing stays consistent.
- Pick MOV when you plan to color grade, key, or composite the result.
Limitations
- Flash tier output is limited to 480p or 720p.
- Single outputs are limited to short clips, up to about 30 seconds.
- Reference stacks are capped at 10 images, 5 videos, and 5 audio clips.
- Reference video and audio clips must stay within per-clip and total-duration limits.
- Providers may reject reference media that contains real human faces.
- A fixed seed helps comparisons, but it won’t guarantee identical frames.
- Low-quality references reduce control. Heavy compression, motion blur, and noisy audio can degrade results.
- Messy or contradictory references can cause identity drift, object swaps, or unstable motion.
Safety & compliance
- Only upload media you have rights to use and transform.
- Avoid prompts that request copyrighted characters, logos, or celebrity likenesses.
- Don’t upload personal or sensitive data in reference media.
- Expect automated moderation. Some prompts or references will be blocked or refused.
- Use watermarking when you need clear disclosure for viewers or platform policies.
Example prompts
Great starting points for seedance v2.5 flash reference.
[Image 1] the paraglider pilot runs down the grassy slope, the red and yellow wing fills with air and lifts him off the ground, and he glides out over the Lauterbrunnen valley past waterfalls and snowy peaks. The camera follows from his side, bright summer morning, wind audio.Video Generation
[Image 1] the golden retriever sprints along the beach and leaps into the waves after a tennis ball in slow motion, water splashing around it, late afternoon sun, waves and happy barking audio.Video Generation
[Image 1] the chef tosses the pan and a flame flares up high in the busy restaurant kitchen, then he plates the dish with tweezers as the camera pushes in, sizzling and kitchen bustle audio.Video Generation
[Image 1] the racing sailboat heels hard and cuts through the choppy water of Sydney Harbour, spray bursting over the bow as the crew leans out, the Sydney Opera House behind. Filmed from a chase boat, wind and water audio.Video Generation
API quick start
Run seedance v2.5 flash reference with a single API call.
POST https://api.wiro.ai/v1/Run/bytedance/seedance-v2-5-flash-reference
{
"prompt": "[Image 1] the paraglider pilot runs down …",
"resolution": "480p",
"duration": "auto",
"generateAudio": "false"
}curl
curl -X POST "https://api.wiro.ai/v1/Run/bytedance/seedance-v2-5-flash-reference" \
-H "Content-Type: application/json" \
-H "x-api-key: YOUR_WIRO_API_KEY" \
--data-binary @- <<'JSON'
{
"prompt": "[Image 1] the paraglider pilot runs down …",
"resolution": "480p",
"duration": "auto",
"generateAudio": "false"
}
JSONOther models you might like
Explore similar video generation models.