Create video from a prompt, images, or reference clips. Up to 30 seconds at 480p, 720p, or 1080p, with generated sound.
Made with Seedance 2.5
Explore Seedance 2.5 demos. Play with sound to watch the camera, movement, and scene changes.
Long takes
A singer moves from the dressing room onto the stage in one continuous shot.
Camera & movement
Flowing costumes and a circling camera keep a stage performance moving.
Reference to video
A simple 3D reference becomes a colorful animated journey.
Video editing
A side-by-side edit places the same performance in a new scene.
Visual references
Musicians, a singer, and an audience share a coherent concert scene.
Stylized storytelling
A painterly street scene carries its characters through a continuous story.
Creative control
Start with words, a first frame, or references. Shape the shot with controls for length, format, and sound.
01Longer stories
Build a scene up to 30 seconds long, with time for an entrance, an action, and a closing shot. Describe the camera path and key moments so the sequence has a clear direction.
Draft at 480p or 720p, then choose Seedance 2.5 1080p for a more detailed output. The tool shows the credit cost for each setting.
02Generated audio
Describe dialogue, music, ambience, and effects alongside the action. Audio is generated with the picture and is enabled by default; you can turn it off when you plan to add your own soundtrack.
03Multimodal references
Use images for characters and visual style, video for movement, and audio for sound. Seedance 2.5 accepts up to 30 images, 10 videos, and 10 audio files. Refer to them in your prompt with @image1, @video1, and @audio1.
04Edit & extend
Bring a source clip into the reference workflow and describe what should change or continue. For Seedance 2.5 video editing, say what to keep as clearly as what to change. Edit and extend operations follow the source aspect ratio.
The model, explained
Seedance 2.5 is ByteDance’s video generation model. Start with a written scene, animate a still, or guide the result with image, video, and audio references. This workspace brings those inputs together with generation settings, a credit estimate, and your finished video.
Seedance 2.5 requires a completed purchase. Check the credit total before generating; failed tasks are refunded.
Your workflow
Write subject, action, camera, lighting, and sound. Add first/last frames or reference files if identity or motion must stay put.
Choose 4–30 seconds, an aspect ratio, and 480p, 720p, or 1080p. Review the credit estimate, then generate.
Watch the result, download your video, or revise the prompt for another take. Your completed videos stay in your creations.
Compare the settings available in this studio. Seedance 2.5 offers longer clips and more references, with output up to 1080p.
| Feature | Seedance 2.0 on TryH3 | Seedance 2.5 on TryH3 |
|---|---|---|
| Duration | 4–15 seconds | 4–30 seconds; 5 seconds by default |
| Resolution | 480p / 720p / 1080p / 4K | 480p / 720p / 1080p |
| Reference images | 9 | 30 |
| Reference videos | 3, about 15s total | 10, 2–30s each, 30s total |
| Reference audio | 3, 15s total | 10, 30s total; audio-only reference allowed |
| Output | mp4 | mp4 or mov |
| Audio | Generated with the clip | Generated with the clip (default on) |
Use 480p for a quick preview, 720p for a draft, or Seedance 2.5 1080p for more detail. 4K is not available for this model.
Aspect ratios: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9. First/last-frame tasks use adaptive ratio. Edit and extend also use adaptive ratio and follow the source clip.
More files only help when each one has a job.
In the prompt, name the file: keep @image1's face, follow @video1's camera, play @audio1 as the bed.
For photos with real faces, use the asset upload flow before adding them as references.
Choose first/last-frame mode for your starting and ending images, or reference mode for a mix of supporting files.
Low-contrast or tiny face crops drift by later shots. Use a sharp, front-lit still if identity has to survive 20–30 seconds.
Phone-style talking clip or unboxing. Input: one face still plus a 9:16 prompt. Output: 8–15 seconds with synced speech. Keep the face still in the asset library if it is a real person.
9:16, 720p draft then 1080p master. 5–12 seconds unless the joke needs the full 30.
One take up to 30 seconds: product still, optional hand-model clip, voice bed. 16:9 at 1080p for a landing page; 9:16 for ads.
If the last frame is right but the scene stops early, extend instead of rolling a new identity.
Practical answers about output, references, access, and credits.
Use the generator at the top of this page. Sign in, write your prompt, and choose your settings. Seedance 2.5 is already selected. A completed purchase is required to generate; you do not need to set up an API key.
Choose 4–30 seconds, with 5 seconds as the default. Available resolutions are 480p, 720p, and 1080p. This studio does not offer 4K for Seedance 2.5.
Yes. Audio is on by default and built in the same pass as the picture. Turn it off if you will replace the track later.
Up to 30 images, 10 videos, and 10 audio files. Videos and audio are 2–30 seconds each, 30 seconds combined. Name them in the prompt with @image1, @video1, or @audio1.
Seedance 2.5 extends the maximum clip length from 15 to 30 seconds and supports more references, including audio-only references. Its output here goes up to 1080p. Seedance 2.0 is available in the other video tools if you need 4K.
Use an existing clip as a reference and describe the edit or continuation. Keep the subject, setting, and details you want to preserve explicit. Edit and extend operations follow the source aspect ratio; editing also follows the source length.
There is no free Seedance 2.5 tier on TryH3. Credits are billed per second at the rate shown on Generate for this duration and resolution. Paid generations on this model follow your plan. Confirm license terms on the commercial-license page before client work. Unpaid accounts are limited to the free video model, which carries a site watermark.
Upload photos with real faces through the asset library before using them as references. Direct public image URLs are not supported for those photos.
Usually as a starting point. Rewrite duration and reference tags for the 30-second cap, and do not ask for 4K.
Start with a prompt or your own references. Choose your length, format, and audio, then check the credit estimate before you generate.
