Prompts
Featured
Models
Submit
Blog

AhaPrompt

AI Prompts for Images, Video & Audio — Every Model, One Library

Explore

All Prompts
Featured Prompts
Image Prompts
Video Prompts
Audio Prompts
AI Models
Submit a Prompt
Blog

Languages

English
中文
日本語
한국어

Copyright © 2026 AhaPrompt All rights reserved.

Privacy Policy Terms of Service

HomeAI Prompt LibraryAI Video PromptsDots and Codex Agent Workflow Setup

Dots and Codex Agent Workflow Setup

GPT-6 AstraOpenAIvideo

Set up this Dots + Codex workflow for my current project: 1. Have Codex inspect ~/.codex/agents and .codex/agents for explorer, researcher and worker roles. Read the applicable AGENTS.md instructions. Reuse compatible agents; list conflicting instructions and pinned models before proposing replacements. 2. Draft valid TOML files only for missing roles, including name, description and developer_instructions: > explorer: gpt-6-luna, medium effort → trace relevant files, callers and existing tests; no edits > researcher: gpt-6-luna, medium effort → check documentation and API constraints; no edits > worker: gpt-6.1-sol, medium effort → make scoped changes and run targeted tests 3. Draft model = 'gpt-6.1-sol' and model_reasoning_effort = 'high' in ~/.codex/config.toml. Check overrides (project config, active profiles, shell aliases, launch flags, task settings, agents.default_subagent_model and agents.default_subagent_reasoning_effort). Report conflicts without changing them. 4. Draft one workflow rule in AGENTS.md: delegate independent investigations to explorer and researcher, give worker their findings, and return checkpoints before major implementation, after a repeated failure and before completion. Review those checkpoints yourself, then send the same Codex task a revised plan, a targeted investigation or remaining corrections. Show all proposed changes as diffs first. Apply nothing until I approve. After approval, verify the roles load with the intended models. Preserve existing permissions; do not merge or deploy.

Details

Type: AI Video Prompts
Aspect Ratio: 4:5
Duration: 0:30
Updated: Oct 4, 2026

How to use this prompt

1Copy the full prompt with the copy button above.
2Open GPT-6 Astra by OpenAI, or your preferred AI video generator.
3Paste the prompt and keep the camera and motion terms — they steer the cinematography.
4Match the example's settings (4:5 · 0:30), then iterate — small wording changes shift the output a lot.

Frequently asked questions

Which AI model generated this video?

This clip was generated with GPT-6 Astra by OpenAI. The prompt on this page is the exact text used to create it.

What settings does this prompt use?

The example was generated with GPT-6 Astra using: 4:5 · 0:30. Reusing these settings gives the closest match.

How do I get similar results?

Copy the prompt as-is first, then change the subject while keeping the camera, motion, and lighting terms. Video models are sensitive to wording, so expect a few runs.

Can I modify this prompt?

Yes. Prompts work best as starting points — swap scenes, adjust pacing terms, or change the aspect ratio to fit your project.

More GPT-6 Astra prompts

A house cat sits in front of an empty gilded mirror. A full-size lion slowly materializes inside the glass. The magician throws a blue velvet cloth over the entire frame. The cloth drops: the mirror has vanished, and the lion is standing onstage. Final frame: the lion calmly lies down beside the cat.

Build a Chrome extension called Signal Crawler using Manifest V3. The extension should help me find business problems in public web pages. It must work on the page I have open, not silently crawl other sites. CORE EXPERIENCE When I click the extension icon, a small animated crawler appears on the page. It moves between readable text blocks, visually scans them and highlights useful passages. A floating panel shows what it has found in real time. Clicking the icon again removes the crawler, panel, highlights and all changes to the page. WHAT TO ANALYZE Support three source types: 1. Job posts: look for manual, repetitive work, slow workflows and tasks that could be automated. 2. Customer reviews: look for repeated complaints, workarounds, missing features and problems that cost the customer time or money. 3. Product and company pages: look for what the business sells, who it serves and how a proposed fix could fit its existing workflow. Ignore navigation, ads, cookie banners, forms, code blocks and hidden text. Handle pages that load more content while scrolling. THE CRAWLER Draw the crawler on a transparent canvas above the page without blocking clicks or text selection. Give it a compact body, eight jointed legs and visible stepping motion. It should move automatically from one relevant text block to another. When it finds a useful passage, pause briefly, place a foot near it, highlight the passage and send a small animated signal from that passage to the floating panel. Use a restrained visual style: warm yellow for findings, white for the body, dark outlines and a subtle glow. Movement should feel quick and alive, but the text must remain readable. SIGNAL BOARD For each finding, show: - the exact quote - the page title and URL - the type of signal - the problem it suggests - how many similar signals have been found Group related findings into clusters. Do not treat one isolated sentence as a proven market. Make repeated evidence visually more prominent than single mentions. ASTRA ANALYSIS Add an optional GPT Astra analysis step. The user enters their own API key and model name in the extension settings. Store the key locally and make API requests from the background service worker, never from the page. For each cluster, ask Astra to produce: - a plain-English description of the problem - who experiences it - evidence for why it matters - the smallest plausible fix - a short draft offer - what still needs human verification Astra must not invent quotes, companies, prices or customer counts. Every conclusion in the panel should link back to its source passages. PANEL CONTROLS Include Pause, Resume, Analyze, Copy Findings, Export JSON, Clear and Close. The panel should show scanning progress, the number of passages checked, findings collected and clusters formed. Let me open any finding and jump back to its original position on the page. QUALITY REQUIREMENTS The extension must not break scrolling, links, page layout or text selection. It should not send messages, contact companies or publish anything automatically. On Close, remove every canvas element, highlight, listener, observer and timer. Support reduced-motion preferences. Handle API failures by keeping the collected evidence visible. Use plain JavaScript, HTML and CSS. Provide every file needed for a working Manifest V3 extension, then give me clear installation and testing steps.

make a dynamic 15 second motion graphics video that shows off your best work like it's a resume showreel, go all out

Create a complete, polished motion graphics video called “SIGNAL — Turn ideas into motion” Generate a 20-second motion graphics video at 1920×1080 and 120 FPS. Create the entire animation procedurally in Python. Include the complete editable Python source file. Build the animation, render every frame, inspect the output, and export the finished MP4. Do not stop at a storyboard, mockup, explanation, or individual frames. The final result must be a fully rendered motion graphics video. No external visual assets are necessary. Generate all typography, shapes, backgrounds, grids, lines, cards, transitions, and visual effects directly through code. Create a premium futuristic motion identity for a fictional creative technology company called SIGNAL. The visual language should feel like a combination of modern product design, editorial typography, Swiss graphic design, and high-end technology branding. Palette: Near-black: #111214 Warm white: #F4F1E8 Electric lime: #C7FF45 Soft gray: #A7AAA5 Use a clean geometric sans-serif font such as Geist, Inter, Helvetica, or a comparable system font. Typography should be large, confident, and tightly composed. Use generous negative space. Maintain consistent margins and alignment throughout the animation. Avoid unnecessary decoration. Every scene should have one dominant visual idea. The central visual motif is a SIGNAL LINE. The line begins as a tiny pulse. It expands into a waveform. The waveform becomes a grid. The grid becomes cards. The cards collapse into a single line. That line eventually constructs the SIGNAL wordmark. Every transformation should visibly originate from the previous object. Nothing should feel like an unrelated scene replacement. Motion should feel precise, fluid, mechanical, and slightly weighted. Fast transformations should always resolve into readable compositions. 0–2 seconds: Begin with a near-black background. Place a tiny electric-lime horizontal line in the exact center. The line should initially appear almost motionless. A subtle pulse travels through it. As the pulse reaches the center, the line expands horizontally. Reveal large warm-white typography: “IDEAS” Then underneath: “NEED SIGNAL.” Keep the typography left aligned. Small supporting text: “Turn noise into something worth noticing.” The lime signal line remains visible on the right side of the composition. Allow approximately one second of readable hold. 2–4 seconds: The signal line begins oscillating. Transform it smoothly into a clean waveform. The waveform should react rhythmically rather than randomly. Introduce several thin gray guide lines behind it. The waveform stretches across most of the screen. Large typography changes to: “FIND THE PATTERN” Small supporting copy: “Every idea starts as noise.” Animate individual waveform peaks with slight timing offsets. Keep movement controlled. Do not create chaotic audio-visualizer movement. 4–6 seconds: Freeze the waveform momentarily. Extend vertical lines downward from selected waveform peaks. Extend horizontal lines outward. These lines gradually construct a modular grid. The waveform itself should become one of the grid's horizontal lines. Do not remove it and introduce a new grid separately. Once the grid is established, reveal small labels inside several cells: “IDEA” “BUILD” “TEST” “BREAK” “LEARN” “SHIP” Highlight one cell in electric lime. Headline: “STRUCTURE THE NOISE” Hold long enough for the composition to remain readable. 6–9 seconds: Selected grid cells detach from the grid. They become floating rectangular cards. Each card should preserve its original dimensions and position during the beginning of the transformation. Then allow the cards to reorganize around the center. Use approximately six cards. Cards should have restrained shadows and minimal depth. Most cards should be warm white. One card should remain electric lime. Use short labels: “01 / IDEA” “02 / BUILD” “03 / TEST” “04 / REFINE” “05 / SHIP” “06 / REPEAT” The cards move into a clean horizontal sequence. Headline above: “MOVE WITH INTENT” Motion should have slight inertia while remaining controlled. No bouncing. No excessive overshoot. 9–12 seconds: Compress the cards toward the center. As they approach each other, reduce the gaps between them. The cards merge into one long horizontal strip. The strip becomes electric lime. Preserve the same horizontal axis. The strip begins bending smoothly. Transform it into a flowing continuous line that travels across the screen. Switch the background from warm white to near-black through a geometric wipe generated by the moving strip itself. Large warm-white typography appears: “MOMENTUM” Small supporting line: “ONE MOVE CREATES THE NEXT.” The signal line should travel around the typography without obscuring it. Give the line subtle depth using restrained shading. Do not use glow. Do not use neon bloom. 12–14 seconds: The moving line begins drawing geometric letterforms. First construct: S Then: I Then: G Then: N Then: A Then: L Each letter should emerge from the same continuous line. Avoid simply fading the word into existence. The viewer should clearly understand that the moving signal is physically constructing the wordmark. Once complete, display: SIGNAL Centered. Small text underneath: “TURN IDEAS INTO MOTION” Allow the composition to settle. 14–17 seconds: Collapse the SIGNAL wordmark back into the horizontal line. The transformation should reverse naturally. The line travels toward the center. Create three small pulses alon

[Reference Image Definition] Image1 = Red-dressed ethnic-style cartoon girl | Sole character reference Image1 is the only and highest-priority reference for the character's identity, face, hairstyle, clothing, and art style throughout the entire video. Throughout the entire video, strictly maintain the girl from Image1: Recognizable facial features Facial feature proportions Black double braids Red-and-blue braided hair decorations Ethnic-style headpiece Earrings Red ethnic-style long dress Clothing patterns and fur-trimmed decorations Brown short boots Soft, delicate 2D anime illustration texture Cute, lively, with a slightly playful and mischievous vibe Prohibited: Photorealistic rendering 3D rendering Turning into someone else Changing hairstyle Changing clothing Changing art style Image2 = X profile page screenshot | Sole background and layout reference Image2 is the sole reference for the background, UI, typography, and composition of the entire video. Strictly maintain from Image2: Top banner image Position of the circular avatar frame White UI background All buttons Icons Page structure Whitespace Colors Proportions Throughout the entire video, the camera is completely fixed. Any of the following is prohibited for the entire page: Zooming Panning Rotation Perspective changes Recomposition Except for the specified performance, the UI and background must not change. [Overall Requirements] Generate a 15-second, 9:16 vertical, fixed-camera, 2D cartoon-animation-style short video. Throughout the entire video: The background is always the X profile page screenshot from Image2 Only the character and the text being manipulated move The character is initially located inside the circular avatar frame The character's appearance must be replaced with the red-dressed ethnic-style cartoon girl from Image1 The circular avatar frame position and the page itself remain unchanged at all times The overall rhythm is light, clear, with a touch of comedy No dialogue No narration No speech bubbles No additional subtitles No new unrelated characters No new unrelated logos No new unrelated text [Timeline Action Design] [0.0–2.3 seconds] The entire X page remains completely fixed. Only the inside of the circular avatar frame begins to move. The red-dressed girl inside the circular frame first blinks gently, revealing a sweet smile. She slightly lowers her head to adjust her cuffs, then gently lifts the edge of her skirt, as if preparing to make an entrance. Then she looks up at the page outside, glances down at the text content on the page, and finally looks back at the camera, a playful "up to mischief" smirk slowly spreading across her lips. Movements are small, natural, cute, and lively. No exaggerated distortion. No large swinging motions. [2.3–4.0 seconds] The girl places both hands on the edge of the circular avatar frame, treating it as a small entrance, and naturally climbs out from inside it. The sequence of actions is clearly shown as: Both hands grip the frame edge Upper body leans out first One foot steps out first The other foot follows She gently lifts her skirt slightly to prevent the long dress from getting caught The skirt hem briefly gets lightly caught on the frame edge She looks back and gently tugs at the skirt hem The skirt comes out smoothly She lands lightly on the white whitespace area below the X page Upon landing, the character should be light and natural, with the skirt, braids, earrings, and headpiece swaying gently and naturally. Only the character comes out of the avatar frame. The circular avatar frame itself remains in place, undeformed, and does not disappear. The page itself must not undergo any structural changes. [4.0–7.6 seconds] After landing, the girl looks around and notices there is a lot of text on the page. She immediately shows a delighted expression of "I've got a mischievous idea." She begins tearing off sections of text from the page one by one. The operable targets are all readable text / letters / numbers within the page, including: Profile name User ID Bio text Join date Following count Follower count Text inside buttons Post content text Other readable text / letters / numbers She tears off text only 3 times, with clear actions and crisp rhythm: 1st tear She walks to the area near the profile name, pinches an entire section of text with two fingers, and clearly peels it off the page surface like a sticker. The torn-off text is quickly folded twice in her hands and instantly becomes a small paper airplane. She flicks her wrist and sends the paper airplane flying toward the camera. The paper airplane passes through the foreground and flies out of frame. The corresponding original text position immediately becomes blank white. 2nd tear She tears off a section of bio text. This time the motion is more practiced — she folds it into a paper airplane quickly while walking. Then she crisply tosses it out toward the other side of the camera. The paper airplane flies across the foreground of the frame. The corresponding text disappears simultaneously, leaving only blank space. 3rd tear She tears off a larger block of text from the post area. This block is somewhat larger, and she quickly folds it into a slightly bigger paper airplane with both hands. She first does a small "ready to throw" motion, then suddenly throws it toward the camera. The paper airplane rapidly grows larger toward the camera, then flies out of frame. The corresponding original text area is completely transformed into blank white space. After the 3rd throw, she proudly claps her hands lightly. Important requirements Text must be lifted and torn off like paper stickers It cannot simply disappear Paper airplanes must be folded from the torn-off text Each time a paper airplane flies away, the corresponding original text position must genuinely become blank Paper airplanes should not be so large as to block the entire frame Only text may be removed The following elements must be preserved: Top banner illustration Circular avatar frame Image areas All graphic icons Button outlines Dividing lines Thumbnails Other non-text visual elements [7.6–10.4 seconds] She looks around the page and notices there is still a lot of residual text. Her proud smile slightly fades, revealing a hint of "how is there still this much" expression. She lightly places one hand on her hip, then produces a simple broom from behind her. Then she performs 3 brisk screen-clearing motions: 1st sweep A horizontal sweep across the upper part of the page. An entire row of residual text slides away like thin scraps of paper. 2nd sweep A diagonal sweep from the middle of the page downward. The remaining text continues to be gathered and pushed toward one side of the page. 3rd sweep A firm but not overly forceful sweep toward the lower right of the page. The last pile of residual text is swept entirely out of frame. After sweeping, she lightly taps the broom handle on the ground once, as if satisfied with the finish. Important requirements Only text touched by the broom may disappear Text must be gathered and swept away following the broom's motion It cannot suddenly disappear without reason The broom may only appear after the paper airplane actions are complete There is only one broom — it cannot multiply After sweeping, all old text on the page must be completely cleared. Only the following remain on the page: Background image UI graphic icons Button outlines Image elements Large areas of clean white whitespace [10.4–13.7 seconds] She leans the broom gently to one side, then looks down and takes out a thick black marker from her outfit. She pops the cap off with a "snap," turns around, and faces the large area of white whitespace that has been cleared. Then, using the thick black handwritten marker, she writes two lines in the large blank area from the center to the lower part of the page: This place belongs to Sun Keke now! Thank you all for liking it These two lines of text must be clearly, accurately, and completely displayed. No typos. No missing characters. No garbled text. No substitution with other sentences. No addition of any other new text. Text style requirements: Black Thick marker handwritten feel Written out stroke by stroke naturally Large enough Clearly legible Occupying a prominent position in the frame The writing process must clearly show the pen tip advancing — the characters are written out stroke by stroke, not appearing instantly. [13.7–15.0 seconds] After writing, she puts the marker cap back on. She takes a small step back and looks at her "masterpiece" with satisfaction. Then she holds the broom with one hand and the marker with the other, slightly turning to the side. She then looks back over her shoulder toward the camera, revealing a slightly smug, cute, and playful mischievous smile. Not an exaggerated laugh, but a gentle eye-squint, the happy smile after a successful prank. Finally, there is a static freeze frame for approximately 0.7 seconds. In the freeze frame, the following must be clearly visible simultaneously: The two lines of large text on the page: This place belongs to Sun Keke now! Thank you all for liking it The red-dressed girl looking back at the camera with a playful smile The broom The X page background with all text cleared

Do a full animation with sound of anthropic mascote with the newset model of Opus 5.5 fighting against OpenAI mascotte with the new astra models Sol etc like the intro of pokemon fire red and leaf green. Same animations

Set up model routing for this repo: 1. List this month's task types and how long each answer was. Table only, change nothing 2. Anything that ended past 128K goes to Gemini 4 Argon as one call, no chunking > Terminal and shell work stays on Opus 5.5 > Fixes that failed twice go to GPT-6 Astra > Cache the repo context on Argon so every run reuses it 3. Price each task type at Argon's intro rate next to what it costs today 4. Print the routing table with the price column, then wait for my yes before switching anything.

Build a detailed sports car that transforms into a humanoid robot, with an x-ray mode and interactive exploded view

Browse all GPT-6 Astra prompts Explore AI Video Prompts