More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Keyframe Craft for AI Video

    1

    Create start and end frames that animate cleanly and keep the product recognizable.

    $15

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Keyframe Craft for AI Video

    Example session with this skill installed

    I need the start keyframe for shot 2 of a 9:16 skincare ad. The shot: our serum bottle (label reads "LUMA C15") on a marble bathroom counter, then a woman's hand reaches in and picks it up, and the clip ends on the bottle in her hand. Kling will animate it. Arabic hook text goes top-right in post. Which model, what frame, and what do I record?

    • Read your context and instructions
    • Compiled the keyframe craft

    Summary

    Generate

    one start frame on Nano Banana Pro, with the hand

    not yet in frame. Skip the end frame: a held bottle is the documented hand-distortion trigger, so let Kling find the landing and judge the clip.

    1. The start frame

    Moment: the instant before the action. The bottle stands alone on the marble and the hand hasn't entered. A mid-reach start frame either cuts the motion short or forces the model to invent what came before.

    Composition (9:16): bottle in the

    left third, so the RTL reading path runs into it. The

    top-right ~35% stays genuinely empty, a clean plate for the Arabic hook set in post. Nothing important sits in the TikTok right rail or the bottom caption zone.

    Text on the bottle: "LUMA C15" is diegetic, so it

    is generated. Quote it exactly in the prompt, then check it character by character against the real label.

    • Logo: don't name the brand logo in the prompt. Let the product reference image carry it.

    Reference: pass the registered product reference ID. Don't re-describe the bottle in words "because it's faster", because that's where shape and colour drift between shots.

    2. Model choice

    OptionVerdict
    Nano Banana ProUse if your current provider supports the required product reference and label fidelity; check live price and speed
    Nano Banana 2 (Flash)Use when its current reference controls fit the shot
    RecraftUse only if its current reference controls preserve the approved product identity

    3. End frame?

    No. An end frame of a hand gripping the bottle is exactly where fingers distort. If Kling's landing is weak: crop so only a partial hand enters, show the bottle placed rather than held, or build that frame image-to-image from a real photo.

    4. Record (prompts.json)

    {
      "file": "06_frames/shot-02-start.png",
      "prompt": "<exact prompt sent>",
      "model": "<actual-model-id>",
      "resolution": "<actual-output-size>",
      "seed": null,
      "refs": ["<product-reference-id>"],
      "diegetic_text": ["LUMA C15"]
    }
    

    seed stays null unless the call actually used one. Never invent a value.

    Next steps

    Generate, verify the label, check the top-right is truly empty, then hand the frame to your Kling prompt.

    4. Record (prompts.json

    Connects securely to your tools. The creator never sees your data.

    About this skill

    A strong still is not always a good video input. This skill plans the precise moment, composition and identity references for keyframes that a video model can animate. It shows when an end frame adds control, where to leave space for editable text, and what to verify before generation.

    You receive a frame specification, an image prompt brief and a reproducibility record for each shot.

    Use it before Kling, MiniMax or another image-to-video model. You supply the approved product assets and access to your chosen generation tools.

    A product reference image is needed for exact identity. The skill plans and reviews keyframes; it does not render video, guarantee text fidelity, or perform compositing. Model availability and costs can change.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 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

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    Permissions

    No special permissions declared or detected

    Listed1 month ago
    Updated1 month ago

    What's inside

    Frequently Asked Questions

    Keyframe Craft solves the issue of 'un-animatable' images by enforcing technical constraints like safe-area budgeting and mid-action avoidance, ensuring your initial frames are optimized for video model stability rather than just aesthetics.

    The method applies to still-image tools that produce keyframes, including the model families covered in the included matrix. The resulting frame can be used with image-to-video models such as Kling. Check current model names and options in your provider.

    Yes. It covers 9:16 composition: product in the left third for right-to-left readers, top-right space reserved for Arabic type, and keeping key content clear of TikTok's right rail, the Reels bottom third and Snapchat's edges.

    The SKILL.md (start-frame rules, model selection matrix, 9:16/RTL composition, clean-plate vs. diegetic-text rules, hallucination triggers, and a per-frame reproducibility record) plus the model-capabilities table and the cinematic-language reference library.

    It plans and reviews start frames, and helps decide when an end frame is useful. It does not generate the video or perform post-production.

    Drop the folder into your agent's skills directory (Claude Code, Cursor, Codex and others). Then ask it to plan or review a keyframe before you generate. There's nothing to configure and no API keys.

    New in Image & Video