A-cat-with-carrots/OnlyShot

What it solves

OnlyShot is an industrial-grade production pipeline designed to transform a simple idea into a complete, high-quality AI short drama suitable for platforms like Douyin and Hongguo. It addresses the common bottlenecks in AI video production, such as inconsistent visual quality, random compositions, and high costs of iterative video generation, by introducing a structured, multi-phase workflow.

How it works

The project operates as a skill for Claude Code, integrating with the dreamina CLI for asset generation. It follows a rigorous 5-step industrial process:

1. Creation (Phase 1): Performs market intelligence scanning to avoid outdated themes, develops IP briefs, story architectures, and a "setting bible" for characters and scenes. It generates a comprehensive reference (ref) library of 80-150 images to ensure visual consistency.
2. Storyboarding (Phase 1.5): Instead of going straight to video, it generates candidate keyframes (text-to-image). This allows users to confirm compositions at a fraction of the cost of video generation.
3. Production (Phase 2): Automatically selects from four video generation modes (image2video, frames2video, multiframe2video, or multimodal2video) based on the available assets to produce the final clips.
4. Refinement: Implements a "7-move" viral formula (e.g., adversarial openings, 30s explosions, 180s countdowns) to optimize pacing and engagement.
5. Editing: The final clips are then assembled in CapCut with added TTS voiceovers and BGM.

Who it’s for

Content creators, AI filmmakers, and short-drama producers who want to automate the production of viral-style short-form video content while maintaining professional control over visual consistency and narrative pacing.

Highlights

• Cost-Efficient Iteration: Uses a storyboarding phase to reduce modification costs by 18x compared to direct video generation.
• Visual Consistency: Employs an industrial-grade reference library (80-150 images) and a detailed 8-part prompt SOP for character and scene stability.
• Viral-Optimized Pacing: Integrates a specific "7-move" calibration system to ensure high viewer retention based on platform data.
• Multi-Modal Video Generation: Automatically switches between four different generation modes to handle everything from simple movements to complex one-shot sequences.
• Robust Error Handling: Includes a detailed catalog of 17 failure modes and solutions, specifically for the dreamina CLI and content moderation filters.