Long 3D Reels · skill pubblica Open Skill
Long 3D Reels
Halloween Party
Meme
3 1 2
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3487427/2026-09-30/dhqzzaks.png
# Long 3D Reels Create a **30–60 second** vertical (9:16) 3D cinematic Reel: dynamic, rhythmic, visually amazing, entertaining, with many dynamic scenes, smooth flow, realistic cinematic look, and a catchy hook. The main character is a fresh 3D hero generated per idea (not locked to any preset character). Music is included by default; voiceover is optional. ## When to use - User wants a long-form 3D dynamic vertical reel (30–60s) from a one-liner, script, or reference image. - User wants to redo or tweak one stage: rewrite the script, generate the next clip, restyle a character, or mix the episode. ## Steps 1. **Clarify the idea** (one round max): the concept/hook, the 3D hero character, the vibe (e.g. golden-hour desert, cyberpunk city, fantasy arena), and whether they want voiceover. Default: catchy music on, voiceover off unless asked. 2. **Write the script** with `create_document(content=<full script>, name="<title> script", entity_type="scripts")`. Structure it as **5–8 scenes** (each ~5–8s) that together reach 30–60s. Each scene: shot description, camera move, action, and a one-line hook. Keep a clear "peak energy → chill" or escalating arc with a strong opening hook and a memorable payoff. Use `write_todos` to track the stages. 3. **Generate storyboard images** — one per scene with `generate_image` (`model='t2i'`, `orientation='portrait'`, `resolution='1080P'`, `name="scene N - <desc>"`, `entity_type="storyboard_images"`). Style: photorealistic 3D render, cinematic lighting, dynamic composition, shallow depth of field, atmospheric particles, consistent hero character across scenes (reuse the same character description verbatim in every prompt). 4. **Generate video clips** — one per scene with `generate_video` (`model='i2v'`, `input_urls=[<scene image asset_id>]`, `orientation='portrait'`, `duration=5-8`, `name="clip N - <desc>"`, `entity_type="storyboard_videos"`). Prompt each clip with the camera move, action, and rhythm (e.g. slow push-in, whip pan, low-angle hero shot). 5. **Generate the soundtrack** with `generate_music` (`model='music'`, `prompt=<catchy, rhythmic, genre matching the vibe>`, `name="<title> soundtrack"`, `entity_type="music"`). Aim for a driving, rhythmic track that matches the scene arc. 6. **Optional voiceover**: if the user wants narration/hook lines, generate per line with `generate_tts` (`model='tts.design'`, `text=<line>`, `voice_instruction=<English voice description>`, `name="VO - <line>"`, `entity_type="voiceovers"`). 7. **Mix the final reel** with `mix_clips` (`orientation='portrait'`, `name="<title> final reel"`, `entity_type="final_videos"`). Concatenate all clips in order as `videos`; add the soundtrack as `audio` (and voiceover tracks with a `delay` if used). Adjust volumes so music sits under any voiceover. 8. **Reply** in natural language: summarize the reel (title, scenes, runtime, music/VO), and call `set_guidance` to suggest next actions (rewrite script, restyle character, next clip, remix). ## Notes - Keep the hero character description **identical** across all scene prompts for consistency. - All clips are 9:16 portrait; keep `orientation='portrait'` on every generation and on the final mix. - Intermediate storyboard images/videos will appear on the canvas — that's expected. - Report failures honestly; never fabricate URLs. Politely decline prohibited/NSFW content (video generation does not support adult content on any channel).