# Bicep Heart AI — capabilities and model status

This document describes the current product interface. It is not an AI-provider catalog or an inference API specification.

| Capability | Current status |
| --- | --- |
| Single portrait | One image defines the person in the bicep heart clip |
| Local photo preview and removal | Available in the current browser session |
| Phone-to-computer upload | Available through a short-lived QR pairing |
| Public video generation | Signed-in users with sufficient account credits |
| Package checkout | Enabled |
| Generation model | Kie bytedance/seedance-2-mini; API key required |
| Account result | Status polling and playback/link; current task resumes in the same browser session |

The product has one portrait input and one fixed bicep heart template. Single, Five Pack, Ten Pack are package names, not model identifiers.

Generations request 7 seconds, 9:16, 720p, generate_audio=false, with one reference_image_url in reference_image_urls and the fixed reference_video_urls together. The portrait defines appearance. The clip defines a continuous shot: step back, flex one arm, form a small heart against the bicep and hold it with a smile. The prompt excludes the source I / YOU overlay. The server automatically copies the original soundtrack into the generated picture and saves the finished MP4. The page offers playback and download only after this succeeds. Audio/save failures can be retried without another paid generation. Output fidelity is not guaranteed.

The footer AI links open third-party AI interfaces with a suggested evaluation prompt. They do not call an AI API, send photos, automatically submit messages or make those providers part of the generation pipeline. Google AI Mode is labelled as that destination rather than as the Gemini app.

Related: [Product index](https://bicepheartai.net/llms.txt) · [Display prices](https://bicepheartai.net/pricing.md) · [Photo guide](https://bicepheartai.net/blog/bicep-heart-photo-guide)