
Seedance image to video takes one still picture and animates it. You upload a first frame, describe what should change, pick a length and resolution, and generate a short clip with sound.
This guide covers the photo, the prompt and the settings. It also covers one hard limit you should know before you start: photos of real faces.
This guide is independent and not affiliated with ByteDance. We run Seedance 2.0 and Seedance 2.0 Mini in our Seedance video generator through the WaveSpeed API.
At a glance
- Seedance image to video starts from your first frame. You can add an optional last frame.
- Clips run 4 to 15 seconds (BytePlus docs). Our studio offers 4, 8, 12 and 15 seconds.
- Seedance 2.0 series models do not accept reference images that show real human faces.
- Write what changes, not what the picture already shows.
- Draft on Mini, then finish on Seedance 2.0 if the shot works.
What is Seedance image to video?
It is the mode where a picture, not only words, sets the opening of the clip. Seedance 2.0 is a video model from ByteDance's Seed team, officially launched on February 12, 2026. It accepts text, images, audio and video as input (Seed blog).
In Seedance 2.0 image to video, your photo becomes the first frame. The model keeps the subject, colors and layout, then adds motion. If you give a last frame too, the clip travels from the first picture toward the second. BytePlus lists this as its first-and-last-frame mode.
Output is an MP4. Audio is generated with the video by default: background music, ambient sound and voiceover are part of what the Seed team describes. Our studio lets you turn audio off, and the price stays the same either way.
How do you prepare the first and last frame for Seedance image to video?
Treat your photo like the opening shot of a film. A clean frame saves credits. Five checks cover most problems.
- Match the aspect ratio. Seedance supports 16:9, 9:16, 1:1, 4:3, 3:4 and 21:9. In our studio, Seedance image to video can follow the ratio of your photo. If you want a vertical clip, start with a vertical photo.
- Leave room for motion. A subject jammed against the edge has nowhere to go. Keep space in the direction you want the camera or the character to move.
- Use one clear subject. One person, one product or one animal is easier to hold steady than a crowd.
- Start sharp. A well-lit, uncompressed image gives the model more to work with than a blurry screenshot.
- Own the rights. Use your own photos, licensed images or art you made. Do not upload copyrighted characters or someone else's work.
For the optional last frame, pick an image that is a plausible next state of the first one. Same subject, same lighting, new pose or new camera distance. A last frame from a totally different scene gives the model nothing sensible to bridge.
Can you use a photo of a real person?
No, not with Seedance 2.0 series models. BytePlus states it plainly: "Seedance 2.0 series models do not support direct uploads of reference images or videos containing real human faces" (BytePlus ModelArk docs).
This is a policy of the model family, not a quirk of one tool. CapCut said it restricted the ability to make videos from images or videos that contain real faces in its rollout (CapCut newsroom). The Seed team also says that using a real person's likeness as a subject reference requires identity verification or prior legal authorization.
We cannot promise exactly what the upstream service will block on every upload. Our rule is simple: do not upload unauthorized photos of real people, and do not upload protected characters. Plan your Seedance image to video project around subjects that work.
What works well instead:
- Products: a bottle, a shoe, a gadget on a table.
- Scenes: a street, a beach, a kitchen, a landscape.
- Animals: a cat, a horse, a bird.
- Illustrations: your own drawings, paintings or digital art.
- Characters you designed yourself, including AI-generated characters that are not real people.
How do you write a motion prompt?
For Seedance image to video, describe the change, not the picture. If the photo already shows a red mug on a wooden table, saying "a red mug on a wooden table" spends words on something the model can already see.
A strong motion prompt has four parts:
- Subject action: one clear verb. "She turns her head and smiles."
- Camera move: push in, pull back, pan left, orbit, or hold still.
- Environment motion: steam, wind, rain, fabric, flickering light.
- What stays fixed: the outfit, the logo, the color palette.
Here is an example for a product photo of a ceramic mug:
Steam rises slowly from the mug. The camera pushes in gently from a medium shot to a close-up. Soft window light shifts across the table. Keep the mug shape, glaze color and label unchanged.
Notice what is missing: no description of the mug, the table or the room. The image carries all of that.
Keep one main action per clip. If you want audio, say it in plain words, such as "quiet room tone" or a short line of dialogue. BytePlus recommends a prompt of no more than 1,000 English words. Our prompt box accepts up to 3,500 characters, which is a limit of our site, not of the model. Most good Seedance image to video prompts are far shorter. For more wording ideas, see our Seedance 2.0 prompts guide.
Which model, length and resolution should you pick?
For Seedance image to video, pick the shortest, cheapest setting that proves the shot, then step up. Both models share the same limits on length and aspect ratio. Mini is the lower-cost option and Seedance 2.0 is the one BytePlus recommends for the highest generation quality.
Credits per second in our studio. Text to video and Seedance image to video cost the same:
| Model | 480p | 720p | 1080p |
|---|---|---|---|
| Seedance 2.0 Mini | 6 | 8 | 30 |
| Seedance 2.0 | 12 | 14 | 60 |
Every plan includes 1080p. Here is how that adds up:
| Goal | Suggested setting | Credits |
|---|---|---|
| Check if the photo animates at all | Mini, 4s, 480p | 24 |
| Draft a social clip | Mini, 8s, 720p | 64 |
| Finish a keeper | Seedance 2.0, 8s, 720p | 112 |
| Final high-resolution shot | Seedance 2.0, 8s, 1080p | 480 |
Every Seedance image to video test costs credits, and generating needs a subscription. The cheapest test, 4 seconds at 480p on Mini, is 24 credits, and clips from every plan carry no watermark. See our pricing page for plans, or read the Seedance 2.0 pricing breakdown.
On speed, WaveSpeed's own estimates for Seedance image to video are about 131 seconds for Mini and 247 seconds for Seedance 2.0 (Mini page, 2.0 page). These are estimates, not guarantees. Plan for a few minutes per clip.
Not sure which to use? Our Seedance 2.0 Mini guide compares the two in detail.
How do you fix common Seedance image to video problems?
Change one thing per retry. These fixes are general Seedance image to video practice, not test results from us, so try each on your own photo.
| Problem | Likely cause | Try this |
|---|---|---|
| Subject look changes mid-clip | Too much motion asked for | Shorten the clip, ask for one small action |
| Things move that should stay still | Prompt does not say what holds | Add "keep the logo and background unchanged" |
| Look drifts from your photo | Prompt re-describes the scene | Delete scene description, write only motion |
| Wrong camera move | Camera wording is missing or vague | Name one move: "slow push-in" |
| Edges warp or smear | Busy background or fast camera | Simplify the frame or hold the camera still |
| Cropped or reframed oddly | Photo ratio differs from output ratio | Match the photo to the target ratio |
| Stiff, robotic motion | Action is too tiny or too many | One natural action plus one environment cue |
| Last frame looks forced | End image is too different | Use a closer pose, same light and setting |
The Seed team itself says Seedance 2.0 still needs refinement in detail stability, and notes room to improve multi-subject consistency. That is another reason to keep one subject per photo.
How do you do Seedance image to video in our studio?
Here is the full Seedance image to video flow. Setup takes about two minutes.
- Open the generator and choose Image to Video.
- Upload your first frame. Add a last frame if you want a defined ending.
- Write your motion prompt using the four-part pattern above.
- Pick Seedance 2.0 Mini, 4 seconds, 480p for the first test.
- Choose an aspect ratio, or let it follow your photo.
- Leave audio on, or switch it off. The price is the same.
- Generate, review, then change one variable and retry.
- When the shot works, rerun it on Seedance 2.0 at 720p or 1080p.
Our samples below are text-to-video clips, not Seedance image to video. They show the kind of single-subject shot that is also a good fit for a first frame. This anime-style river scene was made on Mini:
This virtual group rooftop teaser was made on Seedance 2.0:
FAQ
Is Seedance image to video free?
Not on our site: Seedance image to video here needs a monthly or annual plan. Our Seedance 2.0 free guide lists the free trials that exist elsewhere and their limits.
Can I animate a photo of myself?
Not as a reference for Seedance 2.0 series models, which do not accept images showing real human faces. Use a product, a scene, an animal, an illustration or a character you designed.
How long can a Seedance image to video clip be?
Seedance 2.0 and Mini both generate 4 to 15 seconds. Our studio offers 4, 8, 12 and 15 seconds.
Does Seedance 2.0 image to video add sound?
Yes. Audio is on by default and can be turned off. Turning it off does not change the credit price.
What file do I get?
An MP4 file with sound, with no watermark on any plan.
Do I need a last frame for Seedance image to video?
No. The first frame is required. The last frame is optional and works best when it is a close, believable next state of the first image.
Can I use any image I find online?
Only images you have the right to use. Commercial use also depends on the terms of the route you use and your plan. CapCut notes that Seedance 2.0 Mini commercial use is allowed only when the terms of your access route allow it (CapCut).
Ready to try Seedance image to video? Open our Seedance studio and run a 4-second Mini test with a product or illustration.
Sources checked Nov 14, 2026: Seed blog, BytePlus ModelArk docs, CapCut newsroom, WaveSpeed model pages.