AI Video, 3D & Prompt Reference Archive: Exploring 4,000+ Opus 5.5 and Astra Cases

Whenever compelling AI-generated videos, games, or 3D scenes appear on social media platform X (formerly Twitter), creators often wonder which models were used and how the prompts were structured. To address this, creator Cheer Lin (@cheerselflin) built a curated archive gathering prompts, generation outputs, real-world use cases, and original source links in one place, introduced to the community via AI Caffeine (@AI_Caffeine).

Image source: AI 카페인 (@AI_Caffeine) / Cheer Lin (@cheerselflin) via X

Rather than offering an unstructured list of prompts, this archive organizes entries by specific creative tasks, making it easy to locate relevant reference material. Here is a practical overview of how to navigate the repository and utilize these workflows.

1. Four Core Categories and Sorting Options

Cheer Lin's archive divides resources into four functional areas based on task type:

• Image & Video Prompts: Gathers prompts and outputs across posters, product imagery, camera movements, and advertisements.
• Model-Specific Galleries: Groups creations by foundation model, with sorting options by publish date, likes, and view counts.
• Real-World Workflows: Covers workflow steps across coding, 3D, research, and video production.
• Learning Resources: Compiles lectures, tutorials, recommended YouTube channels, and newsletters.

2. Quantitative Breakdown: Opus 5.5, GPT-6 Astra, and Blender-Seedance

The archive catalogs real-world use cases and multi-tool combinations with specific model tallies:

• 2,342 Claude Opus 5.5 Cases: The archive includes 2,342 recorded cases specifically utilizing Claude Opus 5.5.
• 1,813 GPT-6 Astra Cases: Compiles 1,813 documented cases powered by GPT-6 Astra.
• 38 Blender + Seedance Workflows: Organizes 38 production workflows combining 3D tool Blender with the Seedance video generation model.

3. Original Post Tracing and Tool Combination Tracking

Entries in the archive preserve links to original posts alongside the recorded combinations of tools used:

• Creator Account Navigation: When encountering a notable result, users can follow links to the original posts and visit the creator's X account to inspect full project context.
• Tool Combination Visibility: By displaying the combination of tools and models used for each piece, creators can see what specific models can accomplish and assemble suitable toolchains for their own work.
• Prompt and Output Comparison: By viewing prompts alongside generated outputs and practical examples, users can study how effective prompts were constructed.

4. Practical Considerations

Keep the following points in mind when referencing the archive:

• No Korean Localization: The platform currently does not support Korean, prompting AI Caffeine to publicly request Korean language support from the creator.
• External Post Dependencies: Because entries link back to original social media posts, media access may be affected if original creators modify their account privacy or remove their posts.
• Model-Specific Variations: Cases reflect outputs from specified models like Claude Opus 5.5 and GPT-6 Astra; running the same prompts on different models or local setups may produce different results.

Original source

• X (formerly Twitter): AI Caffeine (@AI_Caffeine) Archive Introduction Post
• Cheer Lin Archive: AI Video, 3D, and Prompt Curation Platform