hailuo-2-3/fast/pro/image-to-video
Enhanced 1080p image motion conversion for expressive, fluid video creation
Seedance 2.5 Reference to Video is ByteDance Seed's next-generation approach to turning reference material into a short cinematic clip. It uses your reference images, videos, and audio to drive generation while focusing on more accurate editing, more consistent subjects, and steadier quality when a shot is extended or stitched.
Note: Seedance 2.5 is coming soon. While its public API is being finalized, this page previews the Seedance 2.5 experience, and generation currently runs on Seedance 2.0 Pro. The live input fields, ranges, and limits below reflect the Seedance 2.0 Pro model that powers generation today.
Seedance 2.5 is coming soon; generation currently runs on Seedance 2.0 Pro at these rates, billed per second of generated video:
1) Add your references — Upload reference images under Images (up to 9); they anchor identity, wardrobe, and style for Seedance 2.5 Reference to Video. Add short reference Videos or Audio to guide motion or sound.
2) Describe the shot — Write what should happen and how the camera behaves (subject action, push-in, pan). Chinese ~≤500 characters or English ~≤1000 words is recommended.
3) Respect the limits — Reference videos and audio must be 2–15 seconds each; audio under 15 MB.
4) Choose aspect ratio — Use adaptive for exploration, or a fixed ratio (for example 9:16 or 16:9) for a known deliverable.
5) Set duration — Any integer from 4 to 15 seconds on the current model.
6) Pick resolution — 480p for fast drafts, 720p for a balanced default, 1080p for high detail, or 4K for the sharpest output.
7) Generate audio — Leave enabled for dialogue, SFX, or music; disable it when you only need silent video.
8) Generate and iterate — Swap references, refine the prompt, or adjust ratio, duration, and resolution, then preview the Seedance 2.5 experience.
Enhanced 1080p image motion conversion for expressive, fluid video creation
Animate static portraits with smooth, identity-true motion using Steady Dancer's video-driven generation.
Interpolates start-end frames with refined motion control presets
Animate an image into a smooth 6s video with Hailuo 02 Pro.
LTX 2 retake video modifie the video by the prompt.
Create lifelike speech-synced visuals from scripts or clips with Kling Lipsync for precise facial animation and realistic results.
Seedance 2.5 is coming soon. This page previews the Seedance 2.5 model while its public API is finalized, and generation currently runs on Seedance 2.0 Pro. That means the live input fields and limits reflect Seedance 2.0 Pro until the Seedance 2.5 backend is switched on.
Seedance 2.5 Reference to Video is a multimodal model from ByteDance Seed that turns reference images, short videos, and audio, together with a prompt, into a short cinematic clip. It targets ad creative, film previsualization, character animation, and branded storytelling where reference consistency and native audio matter.
You attach reference material—up to 9 images, 3 short videos, and 3 audio clips—and describe the shot in the prompt. The references anchor identity, wardrobe, style, motion, and sound, while the prompt guides action and camera. On the current model powering this page, only the prompt is strictly required, but references are what make reference-to-video results consistent.
Seedance 2.5 focuses on more accurate video editing, stronger reference consistency, and steadier quality when a clip is extended or stitched. It also aims to follow negative and timestamp-style instructions more reliably, improve multi-language results, and reduce duplicate-character artifacts and stray watermarks.
Yes. When audio generation is enabled (the current default), the model can output video with synchronized speech, SFX, and music, which supports lip-synced clips. You can disable it if you only need silent video. Lip-sync quality still depends on prompt clarity and the scene you describe.
Reference consistency is a core focus of Seedance 2.5. In practice, reference images plus a clear prompt help anchor identity, wardrobe, and tone across frames. On the current model powering this page you can attach reference images, videos, and audio within the supported limits to reinforce consistency.
On the model currently powering this page, prompts are recommended at Chinese ~≤500 characters or English ~≤1000 words. You can attach up to 9 reference images, 3 reference videos (2–15 seconds each), and 3 reference audio clips (2–15 seconds, under 15 MB). Only the prompt is required. Check the live parameter panel, since limits will change once Seedance 2.5 goes live.
Resolution options are 480p, 720p (default), 1080p, and 4K. Aspect ratio can be adaptive (default, where the model picks the closest ratio) or fixed at 16:9, 9:16, 4:3, 3:4, 1:1, or 21:9. Duration is a whole number of seconds from 4 to 15 on the current model, with default 5.
Generations consume credits based on the generated video duration on the model currently powering this page: $0.11 per second for 480p, $0.24 per second for 720p, $0.55 per second for 1080p, and $1.20 per second for 4K. See the Generation section on this page for live pricing, which will update when Seedance 2.5 launches.
RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.
