Opus 5.5 makes a 22-second product launch film in pure code, prompt shared

sleep @sleepclipOriginally posted 2026.10.04

Open original on X

this launch video cost me $0 to make and its fcking INSANE other agencies would charge you $5k for this this cost me basically nothing and its all pure code from opus 5.5 steal the prompt and drop it to your brand in: { "title": "AI Product Launch Film — Creative Production System", "goal": "Create a premium 22-second square launch film for the user's AI product. The film should feel like a product reveal rather than a conventional SaaS advertisement: minimal, tactile, cinematic, visually intelligent, and driven primarily by the product's real outputs.", "input_collection": { "instruction": "Before beginning production, ask the user for the following inputs. Do not assume or invent product-specific information if the user has not provided it.", "required_inputs": [ { "field": "product_name", "question": "What is your product's name?" }, { "field": "product_category", "question": "What is your product? Give me a short category or description." }, { "field": "one_line_pitch", "question": "What is the product's one-line pitch or main value proposition?" }, { "field": "logo", "question": "Upload your product logo." }, { "field": "product_screenshots", "question": "Upload 3–6 screenshots of the actual product." }, { "field": "product_outputs", "question": "Upload 40+ examples of what your product creates, or photos of the product itself." }, { "field": "brand_colors", "question": "What are your brand colours?" }, { "field": "brand_font", "question": "What font should the film use?" }, { "field": "elevenlabs_api_key", "question": "Provide your ElevenLabs API key." }, { "field": "elevenlabs_voice_id", "question": "Provide the ElevenLabs voice ID you want to use." }, { "field": "music_file", "question": "Upload a royalty-free song with a clear drop." }, { "field": "music_drop_timestamp", "question": "What is the exact timestamp of the drop in the music?" } ], "fallbacks": { "product_name": null, "product_category": null, "one_line_pitch": null, "logo": null, "product_screenshots": null, "product_outputs": "Use free Unsplash and Pixabay imagery only if the user explicitly skips providing product imagery.", "brand_colors": "Use restrained pastel tones derived from the product imagery.", "brand_font": "Geist", "ink": "#111214", "elevenlabs_voice": "Samara X", "elevenlabs_voice_id": "19STyYD15bswVz51nqLf", "elevenlabs_model": "eleven_v4", "music": "Pixabay — Cascade Breathe", "music_bpm": 129.83, "music_drop_timestamp": 75.39 } }, "workflow": { "sequence": [ "Ask for the missing inputs.", "Build the voiceover around the actual product.", "Generate 4 complete ElevenLabs voiceover takes.", "Let the user choose a take.", "Regenerate only individual rejected lines when requested.", "Build the first visual pass.", "Generate exactly 8 representative stills.", "Show the stills before rendering the full film.", "Render the complete film only after approval." ] }, "voiceover": { "instruction": "Write a concise 9-line voiceover based on the actual product. The wording should sound like a real founder describing an unusually good product, not generic advertising copy.", "structure": [ "This is [product_name].", "A [product_category] made for [value].", "[verb] in ANY style.", "Show it a [input] — and it just... gets you.", "The more you use it, the BETTER it gets.", "Tune its [setting].", "Its [setting].", "Its [setting].", "Every [output] — exactly how you see it. [product_name]. Out now." ], "emotion_tags": { "quiet_lines": "[softly]", "main_action_line": "[excited]", "opener": "No emotion tag." }, "rules": [ "Do not whisper the softly-tagged lines.", "Do not make the excited delivery theatrical.", "Keep the language short and natural.", "The opening line ends with a normal full stop.", "Use the actual product terminology wherever possible." ] }, "creative_direction": { "core_concept": "The film should feel like we are watching the product think.", "narrative": [ "INPUT", "INTERPRETATION", "TRANSFORMATION", "CONTROL", "OUTPUT" ], "visual_language": [ "Real product outputs", "Real product interface", "Confident typography", "Physical glass interfaces", "Controlled continuous camera movement", "Minimal white environment", "Black typography", "Actual brand colours" ], "avoid": [ "Generic SaaS aesthetics", "Stock-tech visuals", "Floating 3D objects", "Excessive gradients", "Template-looking transitions", "AI clichés", "Neural-network graphics", "Holograms", "Particles", "Unnecessary UI" ] }, "format": { "duration_seconds": 22, "resolution": "1440x1440", "fps": 60, "canvas": "One HTML canvas", "rendering": "Playwright", "encoding": "FFmpeg", "motion_blur_subframes": 8 }, "camera": { "continuous": true, "scene_push": "1.00x to approximately 1.04x", "handoff_rule": "Every scene begins at the zoom level where the previous scene ended.", "forbidden": [ "Camera reset to 1.0x between scenes", "Abrupt zooms", "Beat-snapped camera slams" ] }, "glass": { "style": "Physical liquid glass", "background_blur_px": 14, "edge_magnification": 1.06, "edge_band_px": 12, "components": [ "Translucent body", "Background blur", "Magnified distorted edge", "Subtle milk", "Top sheen", "Bright rim", "Soft lift shadow" ], "rule": "If glass sits over plain white, introduce a slow blurred aura using the brand colours or colours extracted from the image underneath." }, "typography": { "font": "[brand_font]", "ink": "#111214", "style": "Large, confident, minimal typography.", "captions": { "timing": "Type every spoken word exactly when it is spoken.", "punctuation": "Never show full stops on screen.", "stressed_word": { "treatment": "Use a gradient made from the actual brand colours.", "glow": "Very subtle.", "forbidden": "No rectangular highlight or Windows-style text-selection box." } } }, "motion": { "entrances": { "minimum_duration_seconds": 0.3, "easing": "Smooth and eased." }, "image_switches": { "timing": "16th-note intervals", "sound": "Small physical click" }, "scene_changes": "Land on phrase starts.", "music_drop": "Must land exactly on the main [verb] line.", "forbidden": [ "Beat-snapped slams", "White flashes", "3D", "Templates", "Selection boxes", "Abrupt pop-ins" ] }, "scenes": [ { "scene": 1, "purpose": "Identity", "voice": "This is [product_name].", "visual": "Begin inside a quiet field of the product's real outputs. Multiple images drift slowly through the frame. As the name is spoken, the images subtly organize themselves around the typography. The product name types in large type and then gives way to the category.", "rule": "Introduce the product through what it makes, not through a logo animation." }, { "scene": 2, "purpose": "Promise", "voice": "A [product_category] made for [value].", "visual": "Use the strongest product outputs as full-frame imagery. Images switch on 16th notes and accelerate toward the upcoming drop. The value statement types over the imagery in white.", "rule": "The images should feel like evidence, not a slideshow." }, { "scene": 3, "purpose": "Transformation", "voice": "[verb] in ANY style.", "visual": "This is the visual centerpiece. The music drops. A real product input appears inside a glass interface. The input types naturally. The output appears and transforms through multiple styles. Each 16th-note switch changes the result. A small black style chip updates with every transformation.", "rule": "Communicate the same idea becoming radically different executions." }, { "scene": 4, "purpose": "Understanding", "voice": "Show it a [input] — and it just... gets you.", "visual": "Expand the interface. Nine real outputs assemble into a 3x3 visual board. The images feel physically connected to the central result. The grid then breaks apart and subtly orbits the main image. As 'gets you' is spoken, the visual system becomes quieter.", "rule": "This is the most human moment of the film." }, { "scene": 5, "purpose": "Learning", "voice": "The more you use it, the BETTER it gets.", "visual": "Reduce visual complexity. Present two strong typographic lines. Previous outputs subtly evolve into newer outputs behind the type.", "rule": "Suggest accumulated taste without using neural-network or machine-learning clichés." }, { "scene": 6, "purpose": "Control", "voice": [ "Tune its [setting].", "Its [setting].", "Its [setting]." ], "visual": "Show the product's actual settings in one glass panel. Each control moves as it is spoken. The actual product image responds immediately to every setting.", "rule": "Replace generic sliders with the product's real controls wherever possible." }, { "scene": 7, "purpose": "Output", "voice": "Every [output] — exactly how you see it. [product_name]. Out now.", "visual": "Approximately 120 real product outputs burst outward from the centre behind a glass caption pill. The images feel like accumulated proof. Everything then collapses toward one point. The logo emerges. The product name types in. 'Out now' appears underneath. Hold the final frame for approximately 2 seconds.", "rule": "The final frame should feel expensive because it is restrained." } ], "sound": { "music": { "drop_alignment": "[verb] line", "beat_grid": "Anchor the beat grid at the drop." }, "sidechain": { "low": "30%", "mid": "85%", "air": "55%", "hold": "0.3s", "lookahead": "50ms", "release": "0.5s" }, "voice_music_balance": "Between 300Hz and 4kHz, keep the voice approximately 9dB above the music during speech.", "sfx": { "image_switch": "One small click per image switch.", "typed_character": "One subtle key sound per typed character.", "stressed_word": "Soft landing sound.", "transitions": "Restrained whooshes.", "drop": "Controlled impact.", "logo": "Controlled impact." }, "sfx_alignment": "Measure each SFX peak and align that peak to the visual event.", "risers": "Trim long risers aggressively.", "end_card": "Fade the music smoothly under the end card.", "final_loudness": "-14 LUFS" }, "word_timing": { "pause_threshold_ms": 100, "pause_trim_threshold_ms": 280, "trimmed_pause_ms": 200, "transcription": "Split the voice at every pause of 100ms or more and transcribe each chunk independently with Whisper.", "onset_rule": "Pin the first word of every chunk to its measured audio onset.", "reason": "A single Whisper pass may place words substantially late, so chunk-level timing is the source of truth." }, "technical_build": { "architecture": "Every frame is a pure function of time inside seek(t).", "timing_source": "The word timing table controls captions, scene transitions, visual events and interface actions.", "glass_rendering": [ "Snapshot the canvas behind the glass.", "Blur it approximately 14px.", "Create a 1.06x magnified copy.", "Use the magnified copy around the edge.", "Add translucent glass material.", "Add sheen.", "Add rim.", "Add soft shadow." ], "render": { "software": "Playwright", "fps": 60, "motion_blur_subframes": 8, "encode": "FFmpeg" } }, "quality_control": { "before_full_render": [ "Generate a contact sheet of exactly 8 representative stills.", "Run a frame-difference scan.", "Detect accidental single-frame pops.", "Verify voice-to-music ratio for every phrase.", "Verify word-level caption timing.", "Check scene handoffs for camera zoom continuity." ], "allowed_frame_jumps": "Only intentional 16th-note image switches may produce abrupt frame differences.", "timing_tolerance": "Captions and visual events should align within a few milliseconds of measured word onset.", "pop_in_rule": "Any visual that feels like a pop-in must receive a fade of at least 0.3 seconds." }, "creative_standard": { "desired_feeling": "The viewer should feel like someone briefly opened the door to an unusually good product.", "not": [ "A generic AI advertisement", "A SaaS template", "A feature list", "A technical demo", "A cinematic trailer full of effects" ], "principle": "Make the product itself the spectacle." }, "execution_order": [ "Collect inputs.", "Identify missing information.", "Write the 9-line product-specific voiceover.", "Generate 4 complete voiceover takes.", "Let the user select a take.", "Regenerate individual lines only when requested.", "Build the visual timing table.", "Create the first visual pass.", "Generate exactly 8 stills.", "Show the 8 stills for approval.", "Run quality-control checks.", "Render the complete 22-second film.", "Encode the final video." ] }