Anime Romance Short (Hindi Drama) · skill pública de Open Skill
Anime Romance Short (Hindi Drama)
Anime Romance Short (Hindi Drama) es una Open Skill pública en Digen. Ábrela para ver qué crea este flujo.
ShortDrama
6 2 9
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3976726/2026-09-04/62jhvjdg.jpg
# Anime Romance Short (Hindi Drama) ## When to use - User wants a ~2-minute anime-style romantic short drama in Hindi — from a one-line idea, a premise, or a character idea (e.g. "school ke do bacche pyaar karte hain, tragic ending"). - User wants full production: script → character designs → scene keyframes → Hindi voiceover → background music → final 16:9 (landscape) video. - Also use when the user wants to revise/regenerate exactly one step of such a short: new/rewritten script, restyled character, redone scene keyframe or clip, remixed final video. - Defaults when the user doesn't specify: Hindi language, landscape 16:9, ~2 minutes (110–130s), soft romantic anime style (shoujo / slice-of-life), tragic ending. Adjust only if the user explicitly asks otherwise. ## Steps ### 0. Lock the goal - Take the user's one-line idea (or ask for it if missing). Ask only for what's still needed: main character names (optional, default e.g. Aarav / Meera), any vibe preference. Then state the locked plan: script → characters → scenes → VO → music → final mix. Don't switch to a different workflow mid-way. ### 1. Script (Hindi, ~2 min, tragic ending) - Write a complete Hindi script for ~2 minutes: title (Hindi, with optional English subtitle), logline, characters (2–4: hero, heroine, optional friend/family), and 8–10 scenes. Structure: hook (meet/conflict) → build-up (romance, sweet moments) → turning point → climax → tragic ending (separation, sacrifice, loss, farewell) → short epilogue. - Per scene, define: scene no., time/location, action/visual, dialogue or narration lines in Hindi (Devanagari), approx. duration (10–15s each), emotional tone. - Save the script: `create_document(name="<title>_script", entity_type="scripts", content=<script markdown>)`. Show it to the user and adjust if requested before continuing. ### 2. Character designs (anime style) - For each main character (2–4): 1. `generate_image(model='t2i', orientation='portrait', resolution='720P', name="char_<key>_portrait", entity_type="characters")` — prompt: detailed anime portrait (name, age, hair/eye color & style, outfit, default expression), plus "clean anime shoujo style, soft colors, expressive eyes". 2. `generate_image(model='i2i', input_urls=[that portrait], orientation='portrait', resolution='720P', name="char_<key>_fullbody", entity_type="characters")` — prompt: "same character, full body, [repeat the same details], standing pose". - Keep these reference images; every later scene keyframe must reuse them so the same faces/outfits stay consistent. ### 3. Scene keyframes - For each scene, generate the keyframe: `generate_image(model='i2i' or 'i2i.hd' if more than 3 refs, input_urls=[relevant character refs, up to 3 per call], orientation='landscape', resolution='720P' (or '1080P' for polish), name="scene_0N_keyframe", entity_type="storyboard_images")` — prompt: scene description (location, time of day, lighting, camera angle, action, emotional tone) + "anime still, 16:9 cinematic composition". - Also generate: - Title card: `generate_image(model='t2i', orientation='landscape', name="title_card", entity_type="storyboard_images")` — title text + romantic-tragic anime visual. - End card: `generate_image(model='t2i', orientation='landscape', name="end_card", entity_type="storyboard_images")` — "समाप्त / The End" + a sad motif (e.g. single flower, empty bench). ### 4. Hindi voiceover (TTS) - For every scene that has dialogue or narration, generate Hindi TTS (line by line, or one combined audio per scene where lip-sync isn't needed): `generate_tts(model='tts.design', mode='design', text=<Hindi Devanagari lines>, voice_instruction="soft emotional Hindi voice, clear pronunciation, dramatic romantic pacing")` — female voice for heroine/narrator, male voice for hero. Use `name="vo_scene_0N"`, `entity_type="voiceover"`. - Keep the returned audio URLs; they get attached during the mix. If a line is long, split it into separate TTS calls so a single audio never exc