Office Comedy · skill pública de Open Skill
Office Comedy
About programic man in office with his child
Programmatic
4 3 0
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4239354/2026-10-14/8gv7ns24.jpg
# Ahmad Office Comedy A recurring vertical (9:16) short-video format: Ahmad (40, 176 cm, 75 kg) sits in a gaming chair in his warm digital office, surrounded by many monitors and several PC towers; his daughter Helma (9) sits on his right on a stool and his son Hami (4) on his left, both smiling. The clip opens on audience applause, Ahmad delivers his welcome + joke lines in his own voice, audience laughter carries it, and it ends with him saying "follow me" as the camera pulls back. ## Preset assets (already bound to this skill) - `ahmad_photo` — reference photo of Ahmad (identity + styling). - `helma_photo` — reference photo of Helma (daughter, 9). - `hami_photo` — reference photo of Hami (son, 4). - `ahmad_voice` — Ahmad's recorded voice (Persian), used as the driving audio. Pass these keys directly in `input_urls` / `audio_url`; the platform resolves them to the bound asset. If a key is missing, ask the user to upload that asset again. ## Reference files - `references/prompts.md` — the exact image / video / music / SFX prompt templates and the default script lines. Read it before generating. ## Steps 1. Use `write_todos` to list the steps below, then update it as you go. 2. Read `references/prompts.md` for the prompt templates and default lines. 3. **Scene image** — `generate_image`, `model='i2i.hd'`, `orientation='portrait'`, `input_urls=[ahmad_photo, helma_photo, hami_photo]`, prompt from the template. `name="Ahmad office scene"`, `entity_type='storyboard_images'`. Show it to the user; if the composition or identities are off, regenerate before continuing. 4. **Talking clip** — `generate_video`, `model='ai2v'` (image + audio → video), `input_urls=[<scene image>]`, `audio_url=ahmad_voice`, `orientation='portrait'`. `name="Ahmad office clip"`, `entity_type='storyboard_videos'`. 5. **Background music** — `generate_music`, `model='music'`, prompt from the template (light instrumental, no lyrics, must not mask the voice). `name="Office comedy bgm"`, `entity_type='music'`. 6. **Applause / laughter SFX** — `generate_music`, `model='music'`, prompt from the template (audience applause + laughter, no music). `name="Audience sfx"`, `entity_type='sfx'`. 7. **Final mix** — `mix_clips`, one clip: `video` = the talking clip (keep its own voice audio at volume 1.0), `audios` = [bgm at ~0.15 volume, audience sfx at ~0.5 with `delay` 0]. `orientation='portrait'`, `name="Ahmad office comedy final"`, `entity_type='final_videos'`. 8. Reply in natural language, then call `set_guidance` with next-step options (redo a step, change a joke, make another episode). ## Notes - Vertical 9:16 by default; only switch to landscape if the user asks. - Keep the voice intelligible: background music volume low (~0.15), never louder than the voice. - If the user wants new lines instead of the recorded voice, use `generate_tts` with `model='tts.clone'` and `reference_audio_url=ahmad_voice` to keep Ahmad's voice, then feed that audio into step 4. - Intermediate assets (scene image, clip, bgm, sfx) all appear on the canvas — that is expected. - Report failures honestly; never fabricate URLs. Decline any request for sexual/adult content — video generation does not support it on any channel.