Video Call with My Boyfriend · عوامی Open Skill
Video Call with My Boyfriend
Video Call with My Boyfriend Digen پر ایک عوامی Open Skill ہے۔ کھولیں اور دیکھیں یہ ورک فلو کیا تخلیق کرتا ہے۔
ShortDrama
13 1 9
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3947739/2026-08-26/396ne9s2.png
# Character Video Call ## When to use - The user uploads two images and wants to generate a "simulated video call" video: the character appears in a video call interface (with call UI, the other party's avatar frame, call status bar), speaking to the camera. - The first image is the character photo (the main subject); the second image is used for the **top-right avatar frame** of the video call interface and must strictly be the second image uploaded by the user. - **The camera is fixed** (simulating a real video call's fixed camera), and the character may make natural movements toward the camera (gestures, nodding, facial expression changes, etc.). - Chat content is decided by the user; the character's speech uses the video model's built-in voice to read the lines; TTS is forbidden. - **The character's speech language must match the user's input language**: the generated speech must be in the exact same language as the user's chat content. - Content of any intensity is allowed (may include adult/NSFW), as decided by the user. ## Steps 1. **Collect inputs**: Two images (required) — First: character image (main subject); Second: top-right avatar frame image (must be the second image uploaded by the user, used for the small avatar window in the top-right corner of the video call interface). Chat content/topic (decided by the user, used as the lines the character speaks; the generated speech must be in the exact same language as this input), orientation (default portrait 9:16 mobile-call style), content intensity preference. If the user has uploaded only one image, remind the user to upload a second one as the top-right avatar frame; do not skip. 2. **Analyze the character**: Call `recognize_image` on the first uploaded character image; record appearance details (hairstyle, eyes, outfit, art style, etc.) to strictly reproduce the art style. 3. **Generate the video call interface frame**: Call `generate_image`, **`model='i2i.hd.lite'` (mandatory — the image generation step must use exactly this model and no other)**, `input_urls=[first character image, second top-right avatar frame image]`, orientation default `portrait` (9:16). Prompt description: the character is in a video call interface (connected state) — top call status bar (call duration/signal/battery), the other party's avatar frame, name label, mute/camera/end-call buttons, character centered facing the camera, art style exactly matching the first uploaded image; the small avatar window in the top-right corner must use the user's second uploaded image. `name="video_call_frame"`, `entity_type="video_call_frames"`. 4. **Show the interface frame and confirm**: Show the generated interface frame to the user and ask whether it is satisfactory. If not, adjust the prompt based on feedback and regenerate until the user confirms, then continue to the next step. 5. **Generate the speaking video** (must use `ref2v`): Call `generate_video`, **`model='ref2v'` (mandatory — the video generation stage must use exactly this channel and no other)**, `input_urls=[interface frame]` (the interface frame is used as the reference image; ref2v treats all inputs as references), orientation matching the interface frame. Unified prompt description: camera fixed (simulating a fixed camera, no pan/tilt/zoom), the character in the video call interface, after connecting, faces the camera and speaks, lines as <user chat content> (read word for word, in the exact same language as the user's input — never switch to another language), lips strictly synced with the built-in voice, natural blinking and slight head movements while speaking, may make natural movements toward the camera (gestures, nodding, facial expression changes, etc.), strictly imitating real video call behavior. `name="video_call_video"`, `entity_type="final_videos"`. 6. Show the final video to the user. ## Notes - **The image generation step (interface frame) must use `i2i.hd.lite` and only `i2i.hd.lite`** — never use bare `i2i` or `i