Novel Promo Video Maker

The Problem It Solves

On short-video platforms, creators promoting novels, web novels, or story excerpts face a core challenge: how to efficiently transform a block of text into a series of coherent, character-consistent, vertically-oriented videos with synchronized narration? Manual production involves separately conceiving scenes, illustrating them (while striving to keep character appearances consistent), recording audio, and meticulously syncing everything—a cumbersome process that struggles to guarantee both quality and consistency.

How the Skill Works: Core Workflow & Capabilities

@beatra-ai/novel-promo-video-maker aims to automate this entire pipeline. It is not a simple "text-to-video" black box but a step-by-step generator following a clear workflow, ensuring each stage is reviewable and adjustable.

1. Intelligent Storyboarding & Character Locking
Given a story text, the skill first plans the sequence. It generates a storyboard table, delineating hook shots, advancing shots, and a cliffhanger or twist ending. A critical step is creating a character roster: for each recurring character, it generates a standalone reference image (artifact). All subsequent storyboard images are then prompted with a reference to this image. This is the key technical method for solving "character consistency" across multiple shots.

2. Precise Calls Based on Model Capability & Real-Time Pricing
The skill uses beatra.models.list to fetch real-time model cards and pricing for text-to-image (for initial character art), image-to-image (for storyboard frames with character references), text-to-speech, and image-to-video capabilities. All cost estimates are based on currently available models and pricing options. Only after the plan is confirmed do paid calls commence:
- beatra.images.generate to create the character reference images.
- beatra.images.transform (with character references) to generate the static frames for each storyboard shot.
- beatra.speech.synthesize to generate the narration audio for each shot.
- beatra.videos.animate to combine each static storyboard frame with its corresponding narration audio into a dynamic video clip. This step requires explicitly selecting a video model that accepts [image, driving_audio] input, avoiding the default auto model.

3. Controlled Delivery & Extension
The final deliverable is a set of video clips arranged in storyboard order. Each clip comes with detailed status, model, dimensions, duration, and usage information. If a shot needs to be longer, it can be extended via a paid beatra.videos.extend call. Every paid operation throughout the process has its own client_request_id for task tracking and fault recovery.

Applicability & Important Notes

• Focus on Finished Product: This skill's goal is to produce video clips with narration. If you only need the storyboard table without actual image, audio, or video generation, use ai-storyboard-generator instead.
• Narration & Characters: It defaults to a single narration voice. For multi-character voice acting, use beatra-ai-voice-studio. Character consistency relies on the quality of the generated reference images.
• Video Model Selection: When calling beatra.videos.animate, you must carefully select a model, rejecting those that do not accept driving_audio, as this will cause generation failures.
• Cost & Confirmation: Image, speech, and video generation are paid operations. Before execution, the skill clearly presents the plan (including storyboard, narration, canvas, call counts per stage, and maximum cost) and requests confirmation. Any modification constitutes a new, separately charged workflow.
• Fault Recovery: After task submission, execution is asynchronous. The skill has task-ID-based polling and recovery logic but aims to avoid unnecessary retries and cost consumption.