Turn a person photo + a product photo + an optional script into a vertical 9:16 UGC-style video ad. Generates a lifestyle hero image (Nano-Banana Pro Edit), then animates it with native audio using Seedance 2.0 VIP image-to-video.
Detected risks:
muapi-ugc-video-factory is a generative-media skill that turns a person photo plus a product photo (and an optional script and environment description) into a vertical 9:16 UGC-style video advertisement with native dialogue audio. It solves the practical marketing problem of producing an authentic, social-media-ready product ad quickly, without a shoot, by orchestrating a three-stage AI pipeline through the muapi CLI.
The workflow runs three sequential steps whose outputs chain together. First, a GPT chat model (temperature 0, ~200 tokens) writes a director-grade ultra-realistic lifestyle photography prompt from the person, product, and environment inputs. Second, the Nano-Banana Pro Edit model fuses the person and product reference images into a single 1K, 9:16 hero photo, which is shown to the user for approval. Third, Seedance 2.0 VIP image-to-video animates that hero photo into a 10-second vertical clip with synced spoken audio, using a defined CFG scale, negative prompt, and a UGC-style prompt built around the user's script. The skill documents input handling (uploading local files via muapi upload, generating placeholders if inputs are missing), tier notes (VIP supports 9:16 and 4-15s durations, tolerates realistic human faces, and offers a faster variant), and authentication via a MUAPI_API_KEY environment variable with a raw-API curl fallback to the muapi endpoint.
It targets marketers, e-commerce sellers, and content creators who want talking-head UGC ads from a single person and product image. A noteworthy consideration is that the pipeline generates a realistic person appearing to speak a scripted line, so it carries synthetic-media/likeness responsibilities: it should be used with imagery the user is authorized to use. Credential handling via an environment variable is standard, and there is no destructive or exfiltrating behavior.
A required person photo and product photo, plus an optional spoken script (defaults to a short sample line) and an optional environment/scene description. If the person or product image is missing, the skill asks the user to upload one or offers to generate placeholders.
Three sequential stages: a GPT model writes a lifestyle-photography prompt, Nano-Banana Pro Edit fuses the person and product into a 1K 9:16 hero image (shown for approval), and Seedance 2.0 VIP image-to-video animates it into a 10-second vertical clip with native dialogue audio.
Through the muapi CLI, which is configured with a MUAPI_API_KEY environment variable (run muapi auth configure first if it is unset). A raw-API fallback sends the key as an x-api-key header to the muapi endpoint. The key is read from the environment, not hardcoded.
VIP tier supports 9:16 and 4-15 second durations, with 10 seconds ideal for a 1-2 sentence script; longer scripts get compressed and words clipped. For multi-shot ads, generate several hero-image variations and animate each independently, since VIP does not do multi-image image-to-video at 9:16 with audio.
Yes. The pipeline produces a realistic person appearing to speak a scripted line, so it should be used only with a person's likeness and product imagery the user is authorized to use; the tool itself is intended for UGC-style product advertising.
Quick Setup:
.claude/skills/Repository
samuraigpt/generative-media-skills