Developer Building Next.js Whiteboard App Seeks Free Vision API to Generate Image Prompts from Drawings
English summary
A developer is building a Next.js whiteboard app using tldraw. Users draw on a canvas, type a prompt, and click Enhance. The app exports the canvas as base64 PNG and sends it to an AI vision model to generate a detailed image prompt. That prompt is then used with Pollinations.ai to generate a refined image. The developer needs a free vision API that accepts base64 image and text prompt, returns text, and works without a credit card. They tried OpenRouter but encountered model routing issues and are seeking alternative reliable free models such as Gemini, LLaVA, or others.
Chinese summary
一名开发者正在构建一个使用tldraw的Next.js白板应用。用户可在画布上绘制，输入提示并点击增强按钮。应用将画布导出为base64 PNG，发送给AI视觉模型，以生成详细的图像提示。该提示随后用于Pollinations.ai生成精修图像。开发者需要免费的视觉API，能够接受base64图像和文本提示并返回文本，且无需信用卡即可使用。他们尝试了OpenRouter但遇到模型路由问题，现寻求替代的免费可靠模型，如Gemini、LLaVA等。
Key points
The developer is creating a Next.js app with tldraw canvas for drawing and generating images via AI.
开发者正在创建一个使用tldraw画布的Next.js应用，用于绘画并通过AI生成图像。
The current pipeline: canvas base64 → vision model for prompt generation → Pollinations.ai for image generation.
当前流程：画布base64 → 视觉模型生成提示 → Pollinations.ai生成图像。
They need a free vision API that accepts image+text, outputs text, and works without credit card.
他们需要一个免费的视觉API，接受图像和文本输入，输出文本，且无需信用卡。
OpenRouter was tried but routed to wrong models, so they seek alternatives like Gemini or LLaVA.
曾尝试OpenRouter但路由到错误的模型，因此寻求像Gemini或LLaVA这样的替代方案。