Swap a product in an existing video: Gemini Omni edit vs Seedance
To replace an object in a finished clip on Sume, use gemini-omni-flash-1.1 with video_url. Seedance reference-to-video makes a new clip from references.

To swap one product for another inside a clip you already have, use gemini-omni-flash-1.1 on the Video Router with video_url and an edit prompt. Seedance and Kling do not edit an existing video on Sume: Seedance builds a new clip from reference images, videos and audio, and kling-3 takes a prompt or start and end frames only.
That difference decides the model before you write a prompt. The routing below comes from Sume's Video Router docs.
Can Seedance 2.5 edit a video I already have?
No. In the Video Router catalog video_to_video is false for every Seedance id and for kling-3. Seedance 2.5 does accept reference_image_urls, reference_video_urls and reference_audio_urls, and it renders a new 4 to 30 second clip guided by them. That can reproduce a look, a motion or a sound, but it is regeneration. The product, the framing and the timing can all change.
Use Seedance references when you want a new clip in the spirit of an old one. Do not expect frame-for-frame preservation.
Which Sume model edits a clip?
gemini-omni-flash-1.1 routes by the shape of the request, so you never pick an endpoint. Sending video_url selects the video_to_video edit capability, and the prompt describes the edit. resolution is optional and defaults to 720p. There is no aspect_ratio or duration on an edit.
video_url is the edit source, not a reference. It cannot be combined with image_url, end_image_url or reference_*_urls, so you cannot send a product photo alongside it. Describe the replacement in words. Billing is the provider list rate times 1.25 per output second by resolution.
How do the options compare?
Pick by whether the original footage must survive.
| Model | Input | Result |
|---|---|---|
| gemini-omni-flash-1.1 | video_url + edit prompt | Edited version of the source clip |
| seedance-2.5 | Reference images, videos or audio | New clip guided by references |
| kling-3 | Prompt or first and last frames | New clip, no references |
| h3-max-recast | Source video + 1 to 4 person photos | People swapped, not objects |
| higgsfield-genjutsu | Source video + 1 to 8 images | Motion transfer; Object Swap not served |
What does the request look like?
This edit request is the one Sume documents, as a script. The source must be a public HTTPS URL.
import os, requests
body = {
"model": "gemini-omni-flash-1.1",
"prompt": "Replace the bottle with an apple. Keep everything else the same.",
"video_url": os.environ["SOURCE_CLIP_URL"],
"resolution": "720p",
"mode": "async",
}
r = requests.post(
"https://api.sume.com/v1/video-router/generate",
json=body,
headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}",
"Idempotency-Key": "edit-apple-001"},
timeout=60,
)
print(r.status_code, r.text[:400])
What are the limits to plan around?
Gemini Omni Flash 1.1 clips are 3 to 10 seconds in the documented envelope, so check the source length before you start. Native audio is always on and generate_audio: false is rejected. Sume's docs do not publish how faithfully an edit preserves untouched regions, so review the output the way you would review any generated clip.
If you need a longer or higher-fidelity result than an edit gives, regenerate with Seedance 2.5 using the original as a reference video, and accept that it is a new clip. See Kling 3.0 vs Seedance 2.0 for the model trade-offs.
Sources
Related posts
More in Models
- Swap the product in a UGC clip with Gemini Omni Flash 1.1 video edit
Keep a UGC clip that works and change only the product: Sume's video router edit mode for gemini-omni-flash-1.1 takes a video_url and a one-line prompt.
- SynthID audio and MP3: does the mark survive Sume's 128 kbps detach?
Google says SynthID audio marks resist noise, MP3 compression and speed changes. Sume audio detach can output 128 kbps MP3; only some sources carry a mark.
- SynthID text watermark: a Gemini script in a Sume avatar video
Google says SynthID marks Gemini app text. If it becomes a Sume avatar script, the video is a new render, and no docs say the mark carries over.
- Veo 3.1 and Veo 3.1 Fast previews end October 22: what replaces them
Google's deprecations page sets October 22, 2026 as the shutdown date for the Veo 3.1 and 3.1 Fast previews. Dates, the replacement id and what Sume lists.
Written by Sume