Veo Shot Prompting
Choose a Google Veo text, first-frame or subject-reference mode, write a testable short shot prompt, and verify the resulting video and audio.
Work from a shot description with a framing, one clear action, desired camera behavior and intended end state. Record target aspect ratio, duration, audio expectations and planned editorial use. Identify the Google product surface, model ID and region because supported fields can differ across a studio, API and model version. Check applicable reference access and request cost before submitting. ## Key Points - [Google video generation prompt guide](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/video-gen-prompt-guide) - [Google Veo generation best practices](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/best-practice) - [Google Veo 3.1 model capabilities](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/veo/3-1-generate) - [Google REST video model input schema](https://docs.cloud.google.com/gemini-enterprise-agent-platform/reference/rest/Shared.Types/VideoGenerationModelInstance)
skilldb get video-production-skills/veo-shot-promptingFull skill: 34 linesVeo Shot Prompting
Work from a shot description with a framing, one clear action, desired camera behavior and intended end state. Record target aspect ratio, duration, audio expectations and planned editorial use. Identify the Google product surface, model ID and region because supported fields can differ across a studio, API and model version. Check applicable reference access and request cost before submitting.
Choose the mode
- Use text-to-video to explore a visual concept without a required source frame. Describe the subject and action with concrete physical detail. Add setting, lighting, composition and camera motion only as needed to make the action readable.
- Use first-frame image-to-video when an approved still should establish the opening composition. Match source framing to the delivery aspect ratio. Describe subject, environmental and camera motion. The still carries the existing wardrobe, lighting and layout.
- Use first-and-last-frame generation when supported and the edit needs a specific visual landing point. Ensure the two frames preserve scale and viewpoint relationships. Ask for one feasible transition between them.
- Use subject-reference video when the supported mode accepts asset images to guide a recurring subject independently of a first-frame composition. Check the model-specific image count, type and duration before building the request. Google's Veo 3.1 model page documents an eight-second constraint for reference image to video on
veo-3.1-generate-001.
Write a prompt that can be tested
Draft the first prompt as a single shot. For example, a text-only prompt could describe a waist-height locked shot of a brass kettle on blue tile, a hand reaching from screen left to set a lid, and steam thinning before the end. For the same action from an approved still, request the hand movement, placement and steam behavior. Use one camera instruction, such as a locked camera or a slow push. Add an audible cue such as a quiet ceramic click only when the model supports audio and the edit needs it.
Prepare a prompt revision and input sheet before sending the first job. Record reference role, file, model, requested duration, aspect ratio, output count and audio settings. In the documented REST schema, referenceImages cannot be combined with image, video or lastFrame. Validate options against the selected surface's documentation. Keep a reviewable preflight for jobs that use paid credits or third-party assets.
Review the video and iterate
Save the returned operation and output identifiers alongside the prompt revision. Inspect opening, midpoint and last frame for composition, subject continuity, camera path and artifacts. Watch the continuous shot for motion and listen to all generated audio if requested. Compare available metadata and crop against the brief. Log rejected takes and change one instruction at a time. If requested controls are unavailable, use supported input roles or construct the missing transition in the edit. A successful API response establishes that a clip was generated. The editorial checks establish whether it meets the brief.
Sources
Install this skill directly: skilldb add video-production-skills
Related Skills
Video Editing
Techniques for editing video content into compelling visual stories. Covers cutting, pacing, transitions, narrative structure in the timeline, and the editorial judgment that transforms raw footage into focused, emotionally engaging sequences.
Video Lighting
Techniques for lighting video productions to create mood, dimension, and focus. Covers three-point setups, practical and motivated lighting, color temperature management, and the craft of making artificial light feel natural while guiding the viewer's eye.
Video Preproduction
Techniques for planning video productions before cameras roll. Covers scripting, storyboarding, scheduling, location scouting, budgeting, and the organizational discipline that prevents expensive problems during production and post-production.
Video Storytelling
Techniques for telling compelling stories through the video medium. Covers narrative structure, emotional arc design, hook creation, and the integration of visual, audio, and editorial elements into cohesive stories that hold attention and move audiences.
Audio for Video
Techniques for capturing, mixing, and mastering audio in video production. Covers microphone selection, on-set recording discipline, post-production mixing, and the critical role sound plays in making video content feel professional and emotionally compelling.
Camera Techniques
Techniques for operating video cameras with intention and control. Covers framing, movement, exposure, focus, and stabilization choices that turn raw footage into deliberate visual storytelling with every shot serving a narrative purpose.