Which Seedance image-to-video model is available here?
First Take Lab currently exposes Seedance 2.5 image-to-video at 480p for five seconds and 111 credits.
Image-to-video workflow
Facts checked August 12, 2026
Seedance image to video: tell the model what moves, not what it can already see
A source image gives Seedance control over subject and composition, but the prompt still has to direct motion. First Take Lab currently exposes the Seedance 2.5 photo path at 480p, five seconds and 111 credits.
Current model
Seedance 2.5
Image-to-video endpoint
Output
480p · 5s
Current fixed settings
Charge
111 credits
Shown before submit
Source
1 image
Uploaded public HTTPS file
A five-step photo-to-video workflow
1. Choose a clear source image
Use a sharp, rights-cleared image with a readable subject and enough frame space for the requested movement.
2. Open image-to-video mode
The current First Take Lab photo path uses Seedance 2.5 at 480p for five seconds and 111 credits.
3. Describe only what changes
State subject motion, environmental motion, camera movement and the intended end state.
4. Keep the first test simple
Use one dominant subject action and one camera move so a failure is diagnosable.
5. Generate and review identity
Check subject consistency, edges, hands, background motion and camera behavior before another run.
Prepare an image the model can animate
- Use an image you own or have permission to animate.
- Prefer one clear subject over a crowded scene for the first run.
- Leave physical room in the frame for the requested subject or camera movement.
- Avoid tiny faces, obscured hands and compression artifacts when consistency matters.
The image-to-video prompt formula
Subject motion + environmental motion + camera motion + timing + end state + constraints.
- Do not spend most of the prompt redescribing clothes, color and composition already present in the image.
- Use verbs: turns, lifts, crosses, ripples, pushes in, orbits.
- Name a final pose or frame state when the motion needs a destination.
- Keep camera movement secondary to subject motion on the first test.
Common image-to-video failures
| Failure | Likely cause | Next test |
|---|---|---|
| Subject changes identity | Source is unclear or action is too large. | Use a clearer image and smaller motion. |
| Background melts | Too many simultaneous environmental changes. | Keep only one background motion. |
| Hands deform | Fine manipulation exceeds the first prompt’s control. | Use a wider shot or simplify hand action. |
| Upload rejected | Unsupported, private or unsafe media URL. | Re-upload a rights-cleared JPG or PNG through the tool. |
Cost before another take
The current photo path charges 111 credits. If the source, action or framing is still uncertain, fix the image and prompt before paying for another generation.
Need a cheaper text draft?
Seedance 1.0 Lite can test prompt language for 18 credits, but it will not preserve the uploaded subject.
See Seedance 1.0Need multimodal 2.0?
The model family supports richer reference inputs, but First Take Lab has not exposed those unverified production modes yet.
See verified 2.0 scopeFrequently asked questions
What should I write in an image-to-video prompt?
Describe subject motion, environmental motion, one camera move, timing, the end state and constraints. Do not simply repeat what the source image shows.
Can I animate a photo of another person?
Only use images you have the right and consent to animate. The service blocks obvious deceptive impersonation and high-risk content.
Why was my image rejected?
Common causes include an unsafe or private URL, unsupported upload, content policy rejection, or a source that the provider cannot process. Failed provider tasks follow the refund path.
Use only images you own or are authorized to animate. First Take Lab does not grant rights to source material and is not affiliated with ByteDance.