Pollinations.AI
Open-source generative AI platform providing a unified API for text, image, audio, and video generation, powering 500+ community apps with accessible AI infrastructure.
At a Glance
About Pollinations.AI
Pollinations.AI is an open-source generative AI platform built by Myceli AI OÜ and based in Berlin, offering a single unified API for text, image, audio, and video generation. The platform is MIT-licensed, community-driven, and reports 10K weekly active developers and 1.5M daily requests across 500+ live apps, according to the homepage.
What It Is
Pollinations.AI is a developer-facing AI infrastructure platform that abstracts away model hosting and GPU management so builders can focus on their applications. The core offering is a unified REST API at gen.pollinations.ai that routes requests to a wide range of models — including Flux, GPT-5, Claude, Gemini, Seedream, and others — for image, text, audio, and video generation. The platform is fully open source under the MIT license, with code, decisions, and roadmap all public on GitHub.
Unified API and Model Access
The platform consolidates multiple generation modalities under one endpoint:
- Image generation — Flux, GPT Image, Seedream, Kontext models
- Video generation — Seedance, Veo (alpha), Wan-Fast with keyframe support
- Audio — Text-to-speech (OpenAI TTS-compatible, 30+ ElevenLabs voices) and speech-to-text with speaker diarization
- Text — Streaming, tool use, structured output, vision, search, and embeddings
- OpenAI-compatible endpoints — drop-in replacement for existing OpenAI integrations
Authentication uses publishable keys (pk_) for client-side use and secret keys (sk_) for server-side, with per-key model scoping available.
Developer Tooling and Ecosystem
Beyond the HTTP API, Pollinations ships several integration surfaces:
polliCLI — installable vianpx @pollinations_ai/cli, supports text, image, audio, and video generation with--jsonoutput and stdin context for agent pipelines- Node.js/browser/React SDK — available as an npm package
- MCP server —
npx @pollinations_ai/mcplets AI assistants like Claude generate images and audio directly via the Model Context Protocol - Pollen credits — a pay-as-you-go credit system ($1 ≈ 1 Pollen) with a "Bring Your Own Pollen" model where users pay from their own wallets, enabling developers to build apps at zero infrastructure cost
Architecture and Infrastructure
The platform runs on a layered architecture: a Cloudflare Worker at gen.pollinations.ai handles API routing and auth, passing requests to an enter.pollinations.ai gateway that dispatches to specialized image and audio services on EC2, GPU VMs for image models, and third-party APIs (ElevenLabs, Scaleway, Deepseek, Azure-hosted LLMs, Cloudflare AI). Image results are cached via Cloudflare Workers and R2 storage. The GitHub README lists infrastructure supporters including AWS Activate, Google Cloud for Startups, NVIDIA Inception, Azure (MS for Startups), Cloudflare, Scaleway, Modal, and Nebius.
Update: Unified API Launch and May 2026 Releases
The platform recently launched gen.pollinations.ai as its new unified endpoint, replacing the older per-modality URLs. Notable recent additions as of May 2026 include:
- Gemini 3.5 Flash (2026-05-19) — live on the text endpoint with reasoning and tool use
- GPT-5.4-Mini (2026-05-15) — available on the free tier with text, images, tools, and streaming
- Uncapped generations (2026-05-18) — removed artificial timeouts on text and image endpoints
- Keyframe video generation (2026-05-17) — pass start/end images to generate transitions across Seedance, Veo, and Wan-Fast
- Speaker diarization (2026-05-17) —
/v1/audio/transcriptionsnow returns structured JSON with speaker IDs and timestamps
The repository shows 4,523 GitHub stars and was last updated on 2026-05-20, indicating active development momentum.
Community and Open-Source Model
The platform explicitly positions itself as community-shaped: developers can submit apps to a public "Greenhouse" directory, contribute via GitHub issues (with a MentatBot AI assistant that can implement requested features), and earn Pollen credits through quests and activity. The Discord community is the primary support channel. Upcoming roadmap items listed on the homepage include Pollinations Login (drop-in auth), App Hosting, App Discovery, and an Ads SDK for developer earnings.
Community Discussions
Be the first to start a conversation about Pollinations.AI
Share your experience with Pollinations.AI, ask questions, or help others learn from your insights.
Pricing
Free Tier
Free credits to start building, with access to models including GPT-5.4-Mini, Mistral, and DeepSeek V4 Flash. Earn extra Pollen via quests.
- Access to free-tier models (GPT-5.4-Mini, Mistral, DeepSeek V4 Flash)
- Free Pollen credits to start
- Earn extra credits via Pollen Quests
- Text, image, audio, and video generation
- OpenAI-compatible endpoints
Pollen Credits
Pay-as-you-go Pollen credits for accessing premium models and higher usage. $1 ≈ 1 Pollen.
- Access to all models including Flux, Claude, Gemini, GPT-5
- No rate limits with secret keys
- Per-key model scoping
- Bring Your Own Pollen for app developers
- User wallet spending caps
- Earnings share for app developers
Capabilities
Key Features
- Unified API for text, image, audio, and video generation
- OpenAI-compatible endpoints
- Image generation with Flux, GPT Image, Seedream, Kontext
- Video generation with Seedance, Veo, Wan-Fast and keyframe support
- Text-to-speech with 30+ voices including ElevenLabs
- Speech-to-text with speaker diarization
- Streaming, tool use, and structured output for text
- Vision, search, and embeddings
- Pollen credits pay-as-you-go system
- Publishable and secret API keys with per-key model scoping
- polli CLI for humans and AI agents
- Node.js/browser/React SDK
- MCP server for AI assistants like Claude
- Bring Your Own Pollen model for app developers
- Cloudflare R2 image caching
- Open-source MIT-licensed codebase
- Community app showcase (Greenhouse directory)
