HackerNews中文版

launchreel.dev作为一名专注于产品开发的独立创始人，进行内容创作以推广产品并不容易。制作视频意味着需要进行屏幕截图、撰写字幕、将文本转换为配音，并将这些元素组装成画面，这个过程耗时数小时。由于这是一个重复性的工作，逐渐成为了瓶颈。因此，我开发了launchreel。它的工作原理如下：你只需粘贴一个网址（最多3个，外加最多6张图片/视频）。它会应用视觉API提取视觉参考、文本，并利用大型语言模型（LLM）编织故事。它会撰写字幕和配音文本。LLM会提供视频计划，包括画面视觉效果、画面时长、字幕和配音文本。然后，一个人工智能代理将这些计划组装成视频，并添加音乐、配音和字幕。它会调整视频尺寸以适应不同平台——HackerNews、X、ProductHunt、博客、Instagram、Facebook和IndieHackers。它还会撰写帖子内容，并提供可下载的视频。它遵循每个平台的规范（例如，HackerNews不允许使用夸张词汇），但无法保证符合所有的社区行为规范。配音使用文本转语音（TTS），可以选择男性/女性温暖的声音，或不使用配音。文件限制：图片≤5MB，视频≤25MB，总计≤60MB，时长约30秒至5分钟。希望能收到关于每个平台语气模型的反馈，以及元素提取在你的页面上是否有效。请联系：founder@deepship.dev欢迎免费试用。

查看原文

launchreel.devBeing a solo founder focused on building products, content creation to evangelize was not easy. Making videos meant taking screen captures, writing captions, converting text to voice overs and assembling them into frames took hours. This being a repetitive exercise, it was becoming a bottleneck.So I built launchreel.How it works:you paste a URL (up to 3, plus up to 6 images/videos). It applies vision apis to extract visual references, text and LLM weaves a story. It writes captions and voice over text. LLM gives video plan with frame visuals, length of frame and caption and voice over text. An AI agent then assembles the plan into a video, applies music, voice and captions.it adjusts video dimensions to match the platforms - HackerNews, X, ProductHunt, Blog, Instagram, Facebook and IndieHackers. It writes post content as well and video downloadable.It follows each platform's conventions (e.g. no hype words for HN) but can't guarantee mod behaviour. Voiceover is TTS (male/female warm, or none). Caps: images ≤5MB, videos ≤25MB, 60MB total, ~30s–5min length.Would like feedback on the per-platform tone models and whether the element-extraction holds up on your pages. founder@deepship.devIts free to try.

将您的着陆页转化为社交媒体强大的视觉效果。