- Home
- Skills
- Image & Video
- Gemini Veo Studio
Works with the AI Agents you already use
Gemini Veo Studio
Make short videos with Google Veo from your agent: a template panel, a shot list you approve, clips with sound, and finishing in code.
Free
Gemini Veo Studio
Example session with this skill installed
Make a 20-second vertical video of my cat Milo saying hi to my followers. Here's a photo of him: milo.jpg
- Read your context and instructions
- Compiled the gemini veo studio
- Generated the image
Plan: Milo says hi (nothing generated or billed yet)
Template Creators → Talking pet, vertical 9:16, 720p, 3 shots × 8 s (about 24 s).
| Shot | Camera | What happens | Sound |
|---|---|---|---|
| 1 | Locked-off medium shot | Milo on a light grey sofa looks into the camera: "Hi everyone, it's Milo!" | Clear, friendly voice |
| 2 | Slow push-in | Milo stretches, then tilts his head at the camera | Soft purr, cushion rustle |
| 3 | Locked-off close-up | Milo blinks slowly: "See you next time!" | Voice, then quiet room tone |
milo.jpg goes in as a reference image on every shot, so Milo looks the same in all three (reference images need 8 s shots). Every prompt ends with "No on-screen text"; captions are added in the edit.
Shot 1, the exact request (python3 scripts/veo.py --template talking-pet --set subject="a ginger tabby cat named Milo" --set line="Hi everyone, it's Milo!" --aspect 9:16 --ref milo.jpg -o milo-shot1.mp4 --dry-run):
{
"model": "veo-3.1-generate-preview",
"prompt": "Locked-off medium shot in a bright, cozy living room. a ginger tabby cat named Milo sits on a light grey sofa and talks straight to the camera in a clear, friendly voice: \"Hi everyone, it's Milo!\". Soft window light, shallow depth of field, natural fur detail, cinematic. No on-screen text.",
"aspect": "9:16",
"resolution": "720p",
"seconds": 8,
"first": null,
"last": null,
"refs": ["milo.jpg"],
"extend": null
}
Cost: Google bills your Gemini API key per second of video: 3 clips × 8 s = 24 s, plus any retry. Current prices: https://ai.google.dev/gemini-api/docs/pricing
Next: say "go" and I'll generate shot 1 (about 1 to 6 minutes), look at it, then shots 2 and 3. Then I join them with ffmpeg, keep Veo's audio, add your captions and send you milo.mp4.
milo.jpg
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
Make short videos with Google Veo through the official Gemini API, from the first idea to a finished MP4. Your agent plans and finishes; Veo films.
What it does
- Starter panel. A gallery of 15 video templates in four groups (Creators, Brand, Restyle, Celebration), like the Videos page in the Gemini app. Pick one, or describe your own idea.
- Shot list first. The agent writes 3 to 5 shots (4, 6 or 8 s each) with camera, light, dialogue and sound, and shows you the list before anything is generated or billed.
- Clips with sound. One Veo call per shot: text-to-video, first and last frame, up to 3 reference images to keep a character or product the same, and extending a clip. The agent looks at each clip and redoes a bad one (each retry is billed).
- Finishing in code (optional). Join the shots, keep Veo's own audio, swap a stand-in for your real mascot, put a real app UI on a green-screen phone, add captions and an end card, export H.264.
What you need
- A Gemini API key from Google AI Studio, with billing turned on. Veo is not on the Gemini API free tier: Google bills your own account per second of video. The skill itself is free. Current prices: https://ai.google.dev/gemini-api/docs/pricing
- An agent that can run commands in a shell, Python 3.9+,
pip install google-genai, and ffmpeg for finishing.
Privacy
There is no server. Prompts and any images you pass go from your machine only to Google's Gemini API (generativelanguage.googleapis.com), with your own key. The key is read from GEMINI_API_KEY; the skill never prints, logs or saves it.
Limits
- One call makes one 4 to 8 s clip; longer videos are several shots joined together.
- Veo writes on-screen text badly, so text is added in the edit.
- Google's safety filters apply. In the EU, UK, Switzerland and MENA, only adults can be generated. Blocked clips are not billed.
- Generated videos carry Google's SynthID watermark.
Early release (v1.0.0): the scripts follow Google's documented Gemini API for Veo 3.1. Open source, MIT licence: https://github.com/sabaazdn73/gemini-veo-studio-skill-for-claude. Not affiliated with Google.
Needs a Gemini API key with billing on: Veo is not on the Gemini API free tier, and Google bills per second of video. One call makes one 4 to 8 s clip; reference images need 8 s shots; extending a clip adds about 7 s at 720p. Google keeps generated videos for 2 days, so extend a clip within that time. Veo's on-screen text comes out garbled, so text is added in the edit. Google's safety filters apply, and in the EU, UK, Switzerland and MENA only adults can be generated. Early release (v1.0.0): results vary, so plan a retry or two per shot.
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
4 uses
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 1 month ago
- Free with an Agensi account
Requires
Permissions
Shell: runs scripts/veo.py and scripts/panel.py with python3, and ffmpeg for finishing. Network: only Google's Gemini API (generativelanguage.googleapis.com), with the user's own key. Environment: reads GEMINI_API_KEY (or GOOGLE_API_KEY) and the optional VEO_MODEL; the key is never printed or saved. Files: reads the reference images you pass; writes each clip's .mp4 plus a .json with the request and the video URI, and panel.html.