Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+21 moreClaude CodeCursorCodex CLIMuseDots
Manus
Grok
Grok Bot+18 agentsH3 Prompt Writing
Technical prompt engineering for MiniMax H3 video generation modes including T2VA, I2VA, and Ref2VA.
$3
H3 Prompt Writing
Example session with this skill installed
Rewrite this for MiniMax H3 T2VA: A cyberpunk street at night with neon signs reflecting in puddles. A hover-car zooms past. 6 seconds. Add synthwave music and rainy atmosphere.
- Read your context and instructions
- Compiled the h3 prompt writing
I have rewritten your request into the structured MiniMax H3 T2VA format, ensuring the 6-second duration is reflected in the multimodal description and audio layers.
MiniMax H3 T2VA Prompt
integrated_multimodal_description:
0-2s: The scene opens with a low-angle shot of a rain-slicked cyberpunk alley. Neon signs in vibrant pink and cyan flicker, their glow reflecting in deep puddles on the cracked asphalt. Heavy rain falls vertically, creating ripples in the water.
2-4s: A sleek, matte-black hover-car enters from the left frame, zooming past the camera at high speed. The vehicle's blue thrusters emit a soft glow and kick up a fine mist of rainwater as it passes. The camera pans right to follow the vehicle's trajectory.
4-6s: The hover-car disappears into the distance between towering megastructures. The camera settles on a close-up of a neon sign reflecting in a puddle as the ripples slowly dissipate.
overall_soundscape:
Continuous heavy rainfall hitting metallic surfaces and splashing into puddles. A high-pitched, electric hum follows the hover-car as it zooms past, transitioning into a low-frequency Doppler effect as it recedes.
non_diegetic_music:
A driving synthwave track with a prominent analog bassline and rhythmic arpeggiated synthesizers. The tempo is approximately 100 BPM to match the rainy atmosphere, with a slight crescendo as the hover-car enters the frame.
Next steps:
- Copy the sections above into the MiniMax H3 T2VA prompt fields.
- Ensure your generation settings are set to 6 seconds to match the description.
- If you want a different vehicle color, update the 2-4s segment of the integrated description.
Connects securely to your tools. The creator never sees your data.
About this skill
Writing prompts for high-end video generation models often results in generic outputs or ignored keyframes. This skill provides the precise structure required for MiniMax H3, ensuring that text, images, and video references are correctly integrated into the model's specific multimodal format.
What it does
- Multimodal Formatting translates simple requests into structured
integrated_multimodal_descriptionblocks. - Reference Management handles
<Picture>,<Video>, and<Audio>labels for accurate Ref2VA generation. - Keyframe Alignment bridges the motion path between start and end frames for I2VA, L2VA, and FL2VA modes.
- Soundscape Design generates distinct sections for
overall_soundscapeandnon_diegetic_musicto control audio layers. - Duration Matching synchronizes visual descriptions and camera movements to the target 4–15 second video length.
How it works
- Identify the generation mode based on whether you are using text-only, single images, start/end frames, or full references.
- Apply the H3 schema by mapping your creative intent to the specific field order required (e.g., subject definitions, retention analysis).
- Execute the rewrite to convert abstract scene descriptions into concrete technical prompts covering composition, lighting, and camera physics.
Frameworks & tools
This skill is framework-agnostic and works with any LLM capable of processing structured text references. It is specifically designed for the MiniMax H3 video generation model architecture.
Why this beats prompting it yourself
Standard prompting often ignores the complex interplay between visual descriptions and reference labels in H3. This skill enforces the exact field names and logical flow necessary to ensure the model respects your input assets rather than hallucinating new ones.
Use cases
- Generating consistent characters across multiple video clips using Ref2VA subject definitions.
- Creating precise camera transitions between two specific images using FL2VA.
- Orchestrating layered sound effects and background scores that match visual timing.
- Converting basic scripts into technical prompts that control specific lighting and lens movements.
Known limitations
Supports video durations between 4 and 15 seconds. Rewrites are performed in English, though dialogue and lyrics are preserved in their original language.
Restricted to video durations of 4–15 seconds. Outputs prompts in English except for specific dialogue/lyrics.
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
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 19 days ago
- Software
Git