- Home
- Skills
- Image & Video
- Refer to the video to replicate the required items.
Works with the AI Agents you already use
Refer to the video to replicate the required items.
Swap video objects using a reference image with hand-protection and multi-tier sampling modes.
Free
Refer to the video to replicate the required items.
Example session with this skill installed
Replace the coffee cup in this 10-second mp4 with the uploaded soda can image. Use the 5-frame sampling mode for balanced quality.
- Read your context and instructions
- Compiled the refer to
- Generated the image
I have processed the request for video object replacement. The system analyzed the source video frames against the physical features of the provided soda can reference image. Hand structures and background elements were locked to prevent distortion.
Video Processing Report
| Parameter | Value |
|---|---|
| Source Video | workspace/input_video.mp4 (10s) |
| Reference Image | workspace/target_soda_can.png |
| Sampling Mode | 5 frames (720p Balanced) |
| Frames Processed | 60 frames |
| Hand Integrity Check | Passed (No deformation detected) |
| Compute Consumption | 450 Estimated Credits |
Processing Log:
- Analysis: Identified cylindrical object in hand at coordinates [240, 180] to [310, 450].
- Mapping: Extracting textures, metallic reflectance, and logo placement from
target_soda_can.png. - Inpainting: Applying object replacement across 60 sampled frames with hand-masking enabled.
- Interpolation: Smoothing transitions between sampled frames for consistent motion.
Next steps
- Download the generated
output_replaced.mp4from the workspace. - If the object jitters, re-run with the 3-frame sampling mode for higher temporal density.
- Review the hand boundaries; if minor artifacts appear, increase the negative prompt weight for skin protection.
target_soda_can.png
PNG · 1536×1024
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Manual video editing for object replacement is tedious and frame-heavy. This skill automates the process of swapping specific handheld objects in a video using a reference image rather than complex text prompting. It preserves the surrounding environment and hand structures while mapping the new object's features onto the original motion path.
What it does
- Image-driven replacement Swaps video objects based on physical characteristics of a reference photo instead of text descriptions.
- Multimodal sampling Offers three distinct sampling rates (3, 5, or 10 frames) to balance visual smoothness against compute cost.
- Hand protection Utilizes pre-configured negative prompts to prevent hand deformation and skin distortion during the inpainting process.
- Compute estimation Provides a front-end credit and resource estimate before initiating API calls.
- Resolution control Supports up to 720p output depending on the selected sampling mode.
How it works
- Upload assets Provide a source MP4 video (up to 60 seconds) and a target reference image.
- Configure sampling Select a sampling interval to determine the density of the replacement frames.
- Set API credentials Input a personal Gemini API Key to handle the multimodal processing.
- Review estimate Check the predicted resource consumption and confirm the task execution.
Frameworks & tools
This skill utilizes the Gemini multimodal API for visual analysis and video inpainting. It requires standard web permissions for file uploads and API requests.
Why this beats prompting it yourself
Generic text-to-video tools often struggle to maintain the exact appearance of a specific product or object across frames. This skill uses image-driven conditioning to ensure the replacement matches your specific reference material while enforcing strict negative constraints to keep human hands looking natural.
Use cases
- Product placement Replace generic items in lifestyle videos with specific branded products.
- E-commerce optimization Update outdated products in existing promotional videos without reshooting.
- Rapid prototyping Visualize how different props look in a scene using static reference images.
- Social media content Swap handheld items for viral or creative visual effects.
Known limitations
Videos are limited to a maximum duration of 60 seconds. High-fidelity results are dependent on the resolution of the reference image and the chosen sampling density.
Maximum video length is 60 seconds. Requires a user-provided Gemini API Key. Output resolution is capped at 720p.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 17 days ago
- Free with an Agensi account
Permissions
This skill needs Read Files to receive user uploaded video and reference image, and Network Access to call Gemini multimodal API for frame replacement. No local files will be stored.