Rain Relaxation Video · Open Skill 公開技能
Rain Relaxation Video
relaxed video skill
Beauty
2 0 3
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4167672/2026-10-05/4degmkdg.jpg
# Rain Relaxation Video Create a calming "rain on the window" guided-relaxation video (soothing narration + rain ambience + slow camera over blurred city lights) plus a matching cover image. Defaults: portrait 9:16 (social), ~60 seconds, a fresh calming script generated each time (use the user's own text if they provide one), burned-in subtitles of the narration, and a rain-ambience soundtrack layered under the voiceover. ## When to use - The user wants a rain/relaxation/ASMR-style calming video with narration and visuals. - The user wants a cover/thumbnail image for such a video. - The user wants to redo, tweak (change script, restyle visuals, change voice, change length), or continue one of these videos. ## Steps 1. First call `write_todos` to list the steps below, and update the whole list as you progress. 2. **Narration script.** If the user provided their own narration text, use it verbatim. Otherwise write a short calming script (~40–70 words) in the user's language, in the same spirit as: "Take a deep breath. Listen to the gentle raindrops dancing on the glass. Let every drop wash away your stress and bring calm to your mind." Save it with `create_document(content=<script>, name="Narration script", entity_type="scripts")`. 3. **Voiceover.** Call `generate_tts` with `model='tts.design'`, `text=<script>`, and a `voice_instruction` describing a soft, slow, warm, soothing female/male voice (match the user's language). Use `name="Narration voiceover"`, `entity_type="voiceovers"`. 4. **Rain ambience.** Call `generate_music` with `model='music'` and a prompt for gentle rain falling on a window, soft distant city hum, calm and slow, ambient, no melody, ~60s. Use `name="Rain ambience"`, `entity_type="music"`. 5. **Visual still.** Call `generate_image` with `model='t2i'`, `orientation='portrait'`, `resolution='1080P'`, prompt: rain sliding down a window pane, blurred warm city lights bokeh in the background, cozy night atmosphere, soft cinematic lighting, photorealistic, calm and dreamy. Use `name="Rain window still"`, `entity_type="storyboard_images"`. 6. **Slow camera video.** Call `generate_video` with `model='i2v'`, `input_urls=[<the still from step 5>]`, `orientation='portrait'`, `resolution='1080P'`, `duration` ~10s, prompt: slow gentle camera movement, rain drops sliding down the glass, blurred city lights softly glowing, calm and meditative, no people. Use `name="Rain window clip"`, `entity_type="storyboard_videos"`. 7. **Cover image.** Call `generate_image` with `model='t2i'`, `orientation='portrait'`, `resolution='1080P'`, prompt: a beautiful cover/thumbnail of rain on a window with blurred city lights, calm blue-teal night tones, space at the top for a title, cinematic, high quality. Use `name="Cover image"`, `entity_type="covers"`. 8. **Mix.** Call `mix_clips` with one clip: `video=<the clip from step 6>`, `audios=[<rain ambience from step 4> (volume ~0.5), <voiceover from step 3> (volume ~1.0)]`, and `subtitles.items` = the narration script split into cues timed to the voiceover (use `font_name='Noto Sans CJK SC'` for Chinese/Japanese/Korean text), `orientation='portrait'`. Use `name="Rain relaxation video"`, `entity_type="final_videos"`. If the user wants a longer video, loop/extend the clip and re-time the audio/subtitles accordingly. 9. Reply in natural language summarizing the final video and cover image (don't paste large URL blocks). Call `set_guidance` to suggest next actions (e.g. "change the narration", "make it longer", "restyle the visuals", "redo the cover"). ## Notes - Use the user's language for the narration, subtitles, and replies. - If the user only wants the cover image or only the video, skip the unneeded steps. - Report failures honestly; never fabricate URLs. Video generation does not support NSFW/adult content — politely decline such requests. - Intermediate products (script, voiceover, still, clip) will appear on the canvas alongside the final video and cover; that's expected.