7 Must‑Watch Open‑Source Prompt Libraries for AI Image and Video Generation (2025‑2026)

2025‑2026 AI image‑generation shift from model breakthroughs to prompt‑engineering maturity.

1. YouMind-OpenLab/awesome-nano-banana-pro-prompts

⭐ 11,272 | TypeScript | CC BY 4.0 | Latest push: 2026‑04‑20 | Repo: https://github.com/YouMind-OpenLab/awesome-nano-banana-pro-prompts

Overview

World’s largest Nano Banana Pro prompt collection, >10,000 hand‑picked prompts, each with a preview image, supporting 16 languages (including Simplified/Traditional Chinese, Japanese, Korean). Companion web gallery (youmind.com) provides category browsing and full‑text search.

Structure

Prompts classified by usage scenario and visual style.

Usage scenarios : Avatar, social‑media graphics, YouTube thumbnails, e‑commerce main images, game assets, posters/flyers, App/Web design, etc.

Style categories : Photography, cinematic, Anime/Manga, illustration, 3D rendering, Chibi, watercolor, ink‑style, cyberpunk, minimalism, pixel art, etc.

Key features

Self‑hosted web gallery rendered from Markdown list.

Each entry includes full prompt text ready for copy‑paste into Nano Banana Pro or Gemini.

Active development; push on 2026‑04‑20.

README available in 16 languages.

2. ZeroLu/awesome-nanobanana-pro

⭐ 9,779 | Markdown | CC BY 4.0 | Repo: https://github.com/ZeroLu/awesome-nanobanana-pro

Overview

Independent developer maintains a high‑quality Nano Banana Pro prompt list; daily visits ~20,000. Focuses on depth of prompt engineering rather than quantity.

Content classification

Ten categories: Photorealism & Aesthetics, Creative Experiments, Education & Knowledge, E‑commerce & Virtual Studio, Workplace & Productivity, Photo Editing & Restoration, Interior Design, Social Media & Marketing, Daily Life & Translation, Social Networking & Avatars.

Key features

Each prompt includes fine‑tuned parameters and effect descriptions.

Many prompts presented in JSON format detailing lighting, composition, color, etc.

Sources (X/Twitter, Replicate) cited for traceability.

Independently maintained, offering a purer curation.

3. songguoxs/gpt4o-image-prompts

⭐ 3,315 | JavaScript | Created: 2025‑04‑28 | Latest push: 2026‑01‑08 | Repo: https://github.com/songguoxs/gpt4o-image-prompts

Overview

Comprehensive repository covering GPT‑4o, Nano Banana (NB1 & NB2), and other models. Main advantage is cross‑model coverage, allowing users to consult prompts for multiple generators in a single place.

Content structure

Prompts grouped by model; each section lists model‑specific styles. Includes Chinese prompts for domestic users.

Key features

Cross‑model comparison: same scene presented with prompts for different models.

Rich Chinese prompt collection.

Lightweight Markdown list without a complex gallery system.

4. EvoLinkAI/awesome-gpt-image-2-prompts

⭐ 640 | Markdown (document‑type) | Created: 2026‑04‑18 | Latest push: 2026‑04‑19 | Repo: https://github.com/EvoLinkAI/awesome-gpt-image-2-prompts

Overview

Newly launched project maintained by AI model service provider Evolink, gathering curated prompts and image examples for GPT‑Image‑2.

Content classification

Six categories: Portrait & Photography, Poster & Illustration, Character Design, UI & Social Media Mockup, Community Examples, Comparison (cross‑model testing).

Key features

Fresh community cases contributed after GPT‑Image‑2 release.

Direct Evolink API link provides a domestic‑friendly endpoint without overseas credit card.

Chinese‑friendly README.

Each case cites author’s Twitter handle for traceability.

Special inclusions

Blind‑test comparisons between GPT‑Image‑2 and Nano Banana 2, plus user‑submitted tuning experiences.

5. dongyubin/Awesome-AI-Images-Prompts

⭐ 77 | Markdown (Chinese) | Created: 2025‑10‑23 | Latest push: 2026‑04‑17 | Repo: https://github.com/dongyubin/Awesome-AI-Images-Prompts

Overview

Personal Chinese‑language index covering the widest range of models: Nano Banana Pro/2, 即梦, Seedream 4.0, Hunyuan 3.0, ChatGPT‑4o, Grok Image, FLUX, Ideogram, Imagen, Leonardo AI, Midjourney, Qwen‑Image‑2512, Seedance 2.0, etc. Largest Chinese prompt index.

Content structure

Prompts arranged with models on the horizontal axis and functionality (text‑to‑image / image‑to‑image) on the vertical axis, forming a comparison table. Also lists free‑use tutorials, “white‑label” methods, and API integration guides for each platform.

Key features

Broadest coverage: over ten models.

Maintained entirely in Chinese, reflecting domestic usage habits.

Tool index includes links to free usage methods and paid entry points.

Active updates; last updated 2026‑04‑17.

6. langgptai/awesome-doubao-prompts

⭐ 218 | Markdown (Chinese) | Created: 2025‑06‑17 | Repo: https://github.com/langgptai/awesome-doubao-prompts

Overview

Most comprehensive prompt collection for ByteDance’s Doubao AI, covering writing assistance, programming analysis, creative generation, and image prompts for the “即梦” video model.

Content classification

Divided into application domains: writing assistance, coding problem analysis, creative generation, plus structured prompt templates from official Doubao releases and community contributions.

Key features

Focused on the Doubao ecosystem; most complete Doubao prompt set.

Covers both text and image prompts, bridging language and visual generation.

Chinese‑centric design tailored to domestic user scenarios.

7. songguoxs/awesome-video-prompts

⭐ 491 | Markdown (Chinese) | Created: 2025‑06‑09 | Repo: https://github.com/songguoxs/awesome-video-prompts

Overview

Curated collection of AI video‑generation prompts covering Veo3/Veo3.1, 可灵 (Kling), 即梦 (Doubao video), and 海螺 (MiniMax). Over 50 cases, each with full parameter specifications and descriptive generation results.

Content classification

Indexed by video scenario: unboxing (clay animation, real objects), cinematic (Lord of the Rings, Transformers), advertising (Apple, Nike), ASMR (fruit cutting, cooking), etc.

Key features

Coverage of major domestic video models.

Rich Chinese/Asian cultural scenes, surpassing many English‑centric projects.

Parameterized prompts include shot composition, camera movement, subject description, environment, and sound‑effect notes for reproducible production.

Prompt example (ASMR Kiwi Cutting)

The prompt lists shot composition, camera movement, subject description, environment, and sound‑effect annotations, ready for production use.

Conclusion

Prompt‑engineering has moved from an exploratory phase to an accumulation phase. Nano Banana Pro is the most mature area; YouMind‑OpenLab and ZeroLu together exceed 20 k stars, establishing them as clear leaders. The 即梦/GPT‑Image‑2 space is rapidly growing, exemplified by EvoLinkAI’s April 2026 launch. Video‑generation prompts represent a new blue‑ocean opportunity, with repositories for 可灵, 即梦, and 海螺 still at modest star levels.