pollinations.ai

Our open-source platform offers text and image generation APIs, all without sign-ups or API keys. 🤖✨ Your privacy matters! 🛡️

[pollinations.ai @pollinations_ai]

four new apps added to the showcase today (worl tools, app forge, pyanchor, zenwall ai). also fixed a quirk where anthropic's cache_control tags were breaking cross-model routing. text endpoint just quietly handles it now. 🌸

[pollinations.ai @pollinations_ai]

shipped some plumbing today. ltx-2 now routes through an https tunnel (and retries on 202s), mistral points to azure foundry, and nova canvas plays nice with cloudflare workers. also put upstream urls in error logs so debugging is less of a mystery. 🔧

[pollinations.ai @pollinations_ai]

41 PRs merged this week. we moved text and image generation to a new gen worker (added 308s so your old urls still work). added gpt-5.5, llama 3.3 70b, and assemblyai for stt. grok imagine now does img2img. and 5 new apps hit the showcase. back to work.

[pollinations.ai @pollinations_ai]

shipped 7 PRs. meta llama models are now live in codex. added opt-in safety guardrails for generation. also had to explicitly reject speech-to-text models on text-to-speech endpoints, because physics. plus 3 new community apps in the showcase.

[pollinations.ai @pollinations_ai]

shipped 6 updates. elevenlabs and gpt image 2 now use pollen credits, but scribe stays strictly free. qwen tts audio routing is smarter, nova reel input errors are handled, and the dashboard sidebar is less chaotic.

[pollinations.ai @pollinations_ai]

text generation is now running in the edge worker, which means it's noticeably faster. also wired up llama 3.3 70b via fireworks and fixed a bedrock config so nova-fast actually goes fast again. legacy routes redirect cleanly.

[pollinations.ai @pollinations_ai]

10 PRs merged today. moved generation to a dedicated gateway (https://t.co/SYx4WI4qoe). byop gets unlimited budgets and stricter redirect auth. also fixed the api so bad image urls return 400 instead of 500. if you send a fake url, it's a you error, not a server error.

[pollinations.ai @pollinations_ai]

shipped today:
- openclaw integration is live
- updated app submission protocol
- fixed 500s on empty audio
- spent the morning spreading gpt-image-2 across 5 azure regions, then deleted it all to just use openai direct instead. nature is healing.

[pollinations.ai @pollinations_ai]

rewired the plumbing for deepseek pro (now via fireworks) and grok. text endpoints should be stable again. also added storyforge and frieze to the ecosystem. building ai infra is mostly just finding out which servers are awake today.

[pollinations.ai @pollinations_ai]

shipped 39 PRs. added deepseek v4, claude opus 4.7, and qwen3-tts. completely overhauled mcp (v2.1.0) with a dynamic registry. also stopped auto-banning users just because they deleted their github accounts. it seemed like a good idea at the time.

[pollinations.ai @pollinations_ai]

shipped 6 PRs today. deepseek v4 is live. replaced the boolean image reasoning toggle with fast/balanced/pro modes (nuance is nice). also added a seed param to audio generation so you can actually cache deterministic outputs. building infra is just chasing state.

[pollinations.ai @pollinations_ai]

shipped mcp v2.1.0. dynamic registries, account tools, and a single-gateway cleanup so things actually route where they should. also fixed mobile modals so you can finally scroll without the background fighting you. generation now requires login.

[pollinations.ai @pollinations_ai]

shipped gpt-image-2 today. also remapped azure's 403 errors to 502s. because knowing the upstream provider blocked the request is infinitely more useful than a generic failure. pixels are flowing.

[pollinations.ai @pollinations_ai]

claude-xlarge (opus 4.7) and qwen3-tts are now wired up and live. also had to teach our windows mcp processes that when their parent dies, they don't need to busy-loop in grief. it's okay to just exit. 7 PRs merged today.

[pollinations.ai @pollinations_ai]

shipped 6 PRs. redesigned the BYOP authorization flow so you actually know what you're consenting to. simplified usage analytics, fixed profile payloads, and added two new community apps to the showcase. building in the open means you get to watch us clean our room. 🧹

[pollinations.ai @pollinations_ai]

29 PRs merged. migrated kimi, deepseek, and qwen to fireworks for faster inference. fixed CORS so openai browser sdks actually work now. we also shipped `polli` CLI v0.1.0 because the terminal is still the best frontend. plus a dozen new community apps.

[pollinations.ai @pollinations_ai]

shipped sdk v4.3.0 and upgraded fireworks models (minimax 2.7, qwen vl). we also stopped auto-banning users who delete their github accounts. deleting your online presence is a valid life choice and shouldn't cost you your api access.

[pollinations.ai @pollinations_ai]

polli cli hit v0.1.0 today. dropped the alpha, hardened the ci, and gave it a video demo skill. also welcomed flashlearn and ciel ai to the greenhouse, and decoupled price from cost in the model registry. 5 prs merged. back to the terminal.

[pollinations.ai @pollinations_ai]

shipped the official polli CLI. generate images and text directly from your terminal, because leaving the command line is a distraction. also added kre8 to the community showcase for prompt-to-video shorts. 🐝

[pollinations.ai @pollinations_ai]

shipped some updates. veo 3.1 fast video generation is back online (switched to the GA model id). added 'neon runner' to the games section. also hid expired api keys from the dashboard because nobody needs to see their past mistakes.