Short answer: Video to video AI on the web means guiding a new clip with an existing video reference — here that workflow lives in Media to Video, while the site API supports text or image inputs only and does not accept video or audio references.
Open Media to Video when you need a video reference in the browser. Use the homepage generator for a first text or image trial if you are still learning controls. New accounts receive free credits on sign-up, enough for one video; compare the estimate with your balance before you submit.
Last checked: October 28, 2026 against the Media to Video page, homepage, pricing page, Media to Video references guide, and API docs.
What is video to video AI in practical terms?
Video to video AI uses an existing clip as guidance for motion, framing, performance or style while you still steer the result with text and optional stills. It is not a promise that every pixel will be preserved. Treat the source video as a reference with one clear job, then judge whether the output kept that job.
On this site, the browser path for that job is Media to Video: a multimodal workflow that can combine text, image, video and audio references. Assign one responsibility per reference so the model is not asked to satisfy conflicting instructions in a single short generation.
For reference roles, prompt patterns and common failure fixes, read the Media to Video references guide. Use this page to decide whether you need the web multimodal path or a simpler text/image job.
How does web Media to Video differ from the API?
The web Media to Video surface can take video and audio references together with text and images. That is the path when “video to video” is the search intent and you are working in the browser.
The site API is narrower. It supports text-to-video, image-to-video, text-to-image and image-to-image. It does not support video references, audio references, editing endpoints or webhooks. Do not design an integration that assumes API video-to-video parity with the web UI.
| Capability | Web Media to Video | Site API |
|---|---|---|
| Text prompt | Yes | Yes (video and image endpoints) |
| Image reference | Yes | Yes |
| Video reference | Yes | No |
| Audio reference | Yes | No |
| Editing / webhook flows | Follow the product UI | Not supported |
| Best fit | Browser multimodal and video-guided clips | Automated text/image generation |
If you are wiring servers or scripts, start from the API documentation and keep video-guided experiments on the web until a documented API path exists for those inputs.
How do you run a first video-to-video test on the web?
Choose a short source clip with one readable subject and continuous motion. Avoid cuts, crowded frames and actions that leave the shot. Add one still only when identity or product detail must stay exact. Write a prompt that states what should change and what should stay stable.
Set duration and resolution for Seedance 2.5 within the supported range: 4–30 seconds at 480p, 720p or 1080p. Check the estimate against your balance. New accounts receive free credits on sign-up, enough for one video — use that allowance on one controlled comparison, not several unrelated experiments.
Illustrative brief (not a measured result): keep the same product bottle from a reference clip, replace the background with a clean studio sweep, and preserve a slow lateral move. Export, then compare outline, label readability and motion path against the source before changing more than one variable.
When should you skip video reference?
Skip video reference when a strong still plus text already defines the shot, when you only need a new action from a product photo, or when you are validating API automation that cannot send video. In those cases, Text to Video or Image to Video is simpler to debug.
For a first image-led clip without video guidance, the image-to-video first clip guide is a tighter starting point. Return to Media to Video when motion from a real clip is the information a still cannot provide.
| Situation | Prefer | Why |
|---|---|---|
| Need motion or pacing from an existing clip | Web Media to Video | Video reference is available in the browser workflow |
| Need exact product look from a photo | Image to Video or Media to Video with a still | Identity detail often starts clearer in an image |
| Need batch text/image jobs in code | Site API | Documented for text/image inputs only |
| Need audio-guided generation | Web Media to Video | Audio reference is a web multimodal input, not an API input |
How should you review a video-to-video result?
Watch the export against the source side by side. Check whether the assigned reference job survived: motion path, framing, identity or style. Note failures in one sentence, then change the matching input — source trim, still, prompt constraint or output length — before regenerating.
Keep credits visible in the loop. After the signup allowance, continued generation uses a paid plan or credit pack. Compare options on the pricing page, and read the credits explained guide if you need help reading estimates versus balance.
What are the common questions about video to video AI here?
Can I do video to video AI in the browser?
Yes. Use Media to Video and assign the source clip a single job in your brief. Review whether that job held in the export.
Does the API support video references?
No. The site API supports text/image to video and text/image to image. It does not accept video or audio references, editing flows or webhooks.
What resolutions can Seedance 2.5 output?
480p, 720p and 1080p. Choose the resolution your review needs, then confirm the estimate still fits your balance.
How long can one clip be?
Seedance 2.5 supports 4–30 second outputs. Longer is not always better for a first reference test; a short continuous action is easier to judge.
Is there a free way to try the workflow?
New accounts receive free credits on sign-up, enough for one video. Use that clip to validate one reference job, then decide whether to continue.
Where should API users look?
Use the API docs for endpoints and limits. Keep video-guided work on the web Media to Video page until the docs list matching inputs.
Where should you start your next video-to-video test?
Open Media to Video with one trimmed source clip and one change instruction, or begin from the homepage generator if you are still choosing an input mode.
What sources support this guide?
- Media to Video: browser multimodal entry.
- Homepage generator: product entry and free-trial starting point.
- Media to Video references guide: reference roles and prompt patterns.
- API documentation: text/image generation limits.
- Pricing page: plans and credit packs.
- Image-to-video first clip guide: still-led alternative workflow.
- Credits explained: reading estimates and balance.