AI Drama Video Generator · öffentlicher Open Skill
AI Drama Video Generator
An all-in-one AI workflow to generate engaging drama videos, cinematic story scenes, and realistic dialogues step-by-step.
Film
52 9 42
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3997124/2026-08-27/g6egg5x9.jpg
# Hindi Drama Series Studio ## When to use - User wants a **long AI Hindi drama series** (multi-episode, family drama genre, ~2-minute episodes, 16:9 for YouTube) built from a one-line idea: the skill writes the Hindi storyline + script itself, then generates characters, scenes, Hindi voiceover, music, and stitched episodes. - User wants to **continue** the series with the next episode, revise an episode's script, or redo one scene/clip. ## Fixed user preferences (defaults) - Genre: **family drama** (sukhi-dukhi ghar ki kahaniyan, rishte, emotional moments) - Episode length: **~2 minutes** each (≈ 10–12 scenes of 8–12 seconds) - Orientation: **landscape 16:9**, resolution 720P default (YouTube ke liye) - Voiceover: **Hindi AI voice** with dialogues (generate_tts) - Episode count: **flexible** — ask the user; if they don't specify, propose a plan (e.g. 5–10 episodes) and confirm before building. - Story: the skill writes storyline + full script itself in **Devanagari Hindi**; the user only gives the idea (one-liner). - Communicate with the user in **Hinglish (Romanized Hindi)**; all story/script/voiceover content in Hindi. ## Steps ### 1. Lock scope - Ask/confirm: series title/one-line story, aur kitne episodes chahiye (flexible). If the user is continuing an existing series, first call `memory_query` to fetch the stored series bible (key `hindi-drama-series:<title>`). - Confirm defaults (family drama, ~2-min episodes, 16:9, Hindi VO) unless the user changes them. ### 2. Series bible + episode plan (Hindi) - Write in Devanagari Hindi: title, logline, main characters (naam, umar, rishta, personality, look, signature outfit), setting, and a 1-paragraph summary for each planned episode. - Save: `create_document(content=<Hindi markdown>, name="Series Bible - <Title>", entity_type="scripts", content_type="text/markdown")`. ### 3. Full Hindi script per episode - For the current episode, write scene-by-scene in Devanagari Hindi: scene no., location, time, characters present, camera/action direction, then dialogue lines as separate bullet entries (speaker: dialogue). Keep each scene's dialogue text together so it can be voiced in one TTS call. - Save: `create_document(content=<Hindi markdown>, name="Script - Episode N", entity_type="scripts")`. If the script is very long, split into multiple documents (e.g. "Script - Episode N Part 1"). - To revise an existing script: `get_document(asset_id)` to read it back, then `update_document(asset_id, content=<corrected Hindi markdown>, name=..., entity_type="scripts")` so the same asset is updated. ### 4. Character reference images - For each main character: `generate_image(model='t2i', orientation='landscape', resolution='720P', prompt=<detailed Indian character description: age, face, hair, signature outfit, family-drama look, neutral background>, name="<Character Name> ref", entity_type="characters")`. - Reuse these refs for every scene and episode so faces stay consistent. ### 5. Scene keyframes - For each scene: `generate_image(model='i2i.hd', input_urls=[<character ref images appearing in this scene>], prompt=<scene composition: setting, characters positions, expressions, lighting, emotional tone>, orientation='landscape', resolution='720P', name="Scene N keyframe", entity_type="scene_refs")`. - If a scene has only one character and i2i.hd is slow, bare `i2i` (1–3 refs) is acceptable. ### 6. Hindi voiceover per scene - For each scene with dialogue: `generate_tts(model='tts.design', mode='design', text=<the scene's Hindi dialogues with short pauses between lines>, voice_instruction="Natural Hindi voice, emotional family-drama delivery, clear, medium pace, female/male matching the main speaker", name="VO - Episode N Scene M", entity_type="voiceovers")`. - If the user provides a voice sample to clone, use `generate_tts(model='tts.clone', mode='clone', text=<...>, reference_audio_url=<user's sample>)` instead. ### 7. Background music per episode - `generate_music(prompt="Emotional Hi