“The image is too big to send”
Pick “Images” or search directly and you get three steps: compress to a target size → resize → change the format; in most cases the first step is enough on its own.
Guide
toolun has more than two hundred tools, but nobody opens it thinking “I want to use PDF Merge” — they think “I need to send this contract to the other side, so it has to be merged, page-numbered, watermarked and given a password”. Find a tool by task lists that kind of single job — 37 of them, covering office documents, images, audio and video, development and everyday life — gives each one a one-line goal, and then lays out 3–4 steps, every step linking to a concrete tool on this site and saying which parameters to choose and where the traps are.
Updated 2026-09-092 sources9 min read
toolun has more than two hundred tools, but nobody opens it thinking “I want to use PDF Merge” — they think “I need to send this contract to the other side, so it has to be merged, page-numbered, watermarked and given a password”. Find a tool by task lists that kind of single job — 37 of them, covering office documents, images, audio and video, development and everyday life — gives each one a one-line goal, and then lays out 3–4 steps, every step linking to a concrete tool on this site and saying which parameters to choose and where the traps are.
Two things it is not. It is not an outbound link directory, because every step happens inside toolun, so each one is “click and do it”. And it is not a tool manual: each tool page documents itself, and this page only says what the tool is responsible for inside this chain and how to set it. The “I have a file…” control at the top filters by file type: if you are holding a pile of HEIC files, you only see tasks that can start from HEIC; with no file at all, you see the “from scratch” tasks.
The list grew out of a product audit of searches that returned nothing: “image is too big to send”, “how do I turn a scan into a searchable PDF”, “video is too big for WeChat” — questions that no tool name can match, and exactly what this page exists to catch.
One of the tasks shown after picking “HEIC (iPhone photos)”:
Searching “contract”:
Images › A batch of HEIC → JPG → uniform size → long image
Goal: turn a set of iPhone photos into one long image you can post to WeChat Moments / Xiaohongshu.
1 Image Converter HEIC to JPG in bulk quality around 85 balances size and looks
2 Resize Images uniform width scale to "width 1080", height automatic
3 Long Image Stitch vertical join spacing and background colour optional
4 Image Compress fit the platform limit Xiaohongshu caps one image at 20 MB, WeChat long images are best under 5 MB
Note: when a long image gets very tall some platforms blur it, so split into two beyond 20 photosOffice documents › Contract: merge → page numbers → watermark → encrypt
Word to PDF → PDF Merge → PDF Page Numbers → PDF Watermark → PDF Protect
Office documents › Scanned pages → searchable PDF
Image Converter → Scan Enhance → Image to PDF → PDF OCR
Office documents › Sign a PDF and flatten it
Handwritten Signature → PDF Sign → PDF FlattenThe default shape of a tool site is a catalogue: every tool listed by category, and the user assembles the flow. That works for someone who knows the tool's name, and fails three other kinds of visitor — the one holding a file and unsure which tool fits, the one who knows the job but calls it something else (“remove the background” = cutout, “Apple photos” = HEIC), and the one whose job spans two or three tools. The audit found that searches from all three groups came back empty almost every time. A task page makes the job the entry point: every task is a fixed sequence of tools, and the user needs to know no tool names and to work out no order.
The principle is cut losses first, then process, then shrink. Take “scanned pages → searchable PDF”: convert the format first (many downstream tools cannot read HEIC), then enhance the image (deskewing and shadow removal raise OCR accuracy noticeably, so enhancement has to come before OCR), then assemble the PDF (running OCR over one PDF in a single pass is less work than image by image), and run OCR last. Compression always sits at the very end — it is lossy, and further processing after it magnifies the loss; encryption also comes last, because most tools cannot open an encrypted file. In the signing chain, “flatten” follows the signature, because flattening freezes editable objects into page content that can no longer be changed.
A yellow box flags three kinds of thing: irreversible (redaction, flattening, compression) — keep the original before you run them; legally loaded (an image signature is not the same as a certificate-based digital signature, and an official ID photo has to be taken at a designated studio); and platform rules (WeChat 25 MB, Xiaohongshu (RedNote) 20 MB, a 200 KB application portal) — these are why the task exists, and they decide the parameters of the last step.
The advice on each step is the value that has proved most reliable in that chain, not the only correct one: JPEG quality 85 is the usual turning point between size and looks; a long image 1080 px wide matches the CSS pixel density of mainstream phones; 720p at 1.5 Mbps is the combination that fits two minutes inside 25 MB; a watermark opacity of 15–25% stays visible without covering the body text. Every number can be adjusted on the individual tool page, whose notes explain the reasoning in more detail.
This page is the master index of tasks; each task will later get a flow page of its own (/guide/<slug>) carrying a longer background and screenshots, and a task's id here is the slug of that future page. The task chips in the site-wide search will link to the matching anchor here (#task-<id>), so whichever entrance you come through you land on the same chain.
Pick “Images” or search directly and you get three steps: compress to a target size → resize → change the format; in most cases the first step is enough on its own.
The second task in “Office documents”: Word to PDF → merge → page numbers → watermark → encrypt, with a warning box reminding you to keep an unencrypted copy and send the password through another channel.
“I have a file → HEIC” narrows the list to a few entries, and the first is convert → uniform width → stitch → compress.
The “Development” group: the recipe from the Git cheat sheet + gitignore / licence choice from the developer directory + a mock terminal to rehearse in.
At the moment, yes. The file only ever travels between your browser and your local disk, never through a server; direct hand-off between tools is being built, and once it ships the tasks here become continuous flows.
Because extra steps can cost you (one more compression, one more transcode). A chain is only as long as the minimum set of steps needed to reach the goal.
Try search first (step notes and tags are matched too); if it still comes up empty, use the site-wide search to find a single tool or browse the relevant category. The task list is extended in batches from records of searches that returned nothing.
No. They are the safest values for common cases and can be adjusted on the tool page, whose notes explain the reasoning behind each parameter.
This page is navigation only: the task data loads statically with the page, and search, filtering and bookmarks happen locally in the browser without sending a request. Every tool it links to processes files in the browser (the few marked API call a third-party interface and say so), intermediate results sit in your download folder, and this site's servers never receive a file.
Updated 2026-09-09
37 common in-site workflows for tasks such as searchable scanned PDFs, contract merging and encryption, HEIC conversion, video compression, social covers, and ID photo backgrounds, with 3–4 linked steps each
Pick a file type to show only the tasks that can start from it. Each task is a one-line goal plus 3–4 steps, and every step links to a specific tool with the settings explained; passing files between tools still means downloading and uploading for now (file dock hand-off is on the main roadmap).
把手机拍的合同 / 书页变成能全文搜索、能复制文字的 PDF。
iPhone 照片默认 HEIC,很多下游工具不认;已是 JPG/PNG 可跳过
自动裁边、纠偏、去阴影、提高对比度,OCR 识别率会明显提升
按页码顺序拖放;选 A4 让所有页尺寸一致
生成带隐藏文字层的 PDF:外观不变,但可以搜索与复制
把正文与附件合成一份带页码、有水印、有打开口令的 PDF 发给对方。
全部是 PDF 可跳过
列表顺序就是最终页序,拖动调整
「第 X 页 / 共 Y 页」放页脚居中,封面可从第 2 页起编号
文字水印写「仅供 XX 审阅」+ 日期,透明度 15–25%
打开口令另用其他渠道发给对方;权限口令可禁止打印与复制
把几十 MB 的 PDF 压到邮件附件限制(通常 10–25 MB)以内。
先试「中等」,扫描件类 PDF 通常能减 60–80%
拆成两三份分别发送,或按章节拆
ZIP 对已压缩的 PDF 几乎不再减小,主要是方便打包
拿到 PDF 报告或表格,需要改内容。
能提取出文字说明是「原生 PDF」,转换效果好;提取为空则是扫描件
让它拥有文字层
文档类用这个;排版复杂的页面可能需要手动调整
表格类用这个,逐页识别表格结构
收到需要签字的 PDF,签好后发回,且对方不能再改签名。
鼠标 / 触屏写,导出透明 PNG
可同时填日期与文字
把签名与表单字段固化到页面内容里,不能再被移动或删除
发出去前把身份证号、金额、姓名遮住,且对方无法恢复。
框选区域,工具会删除该区域的文字与图像,不是盖一层黑色
作者、标题、创建软件等字段也可能泄露信息
提取全文,确认被涂掉的内容真的不在了
交给打印店或对方查看时不跑版。
宽表格选横向;先在 Excel 里设好打印区域
讲义类 2 页 / 4 页拼一张
把 Markdown 写的文档交给不用 Markdown 的人。
检查标题层级与代码块
选主题与纸张
转出的 HTML 粘贴进 Word 可保留标题与列表
几十个「扫描001.pdf」按「日期-客户-类型」统一改名。
支持序号、日期、查找替换、大小写;预览确认后再执行
按内容哈希找出重复文件
iPhone 拍的一组照片做成一张可以发朋友圈 / 小红书的长图。
质量 85 左右体积与画质平衡
按「宽度 1080」缩放,高度自动
可加间距与背景色
小红书单图 ≤ 20 MB,微信长图建议 < 5 MB
微信 / 邮件 / 表单上传提示图片超限。
设目标大小(如 2 MB)让工具自动找质量参数
手机屏幕看 1080–1440 宽足够
PNG 截图转 JPG 或 WebP 通常小一半以上
用手机自拍做一张一寸 / 二寸、白底或蓝底的证件照。
模型在浏览器里跑,照片不上传
选一寸 / 二寸 / 各类考试规格,选底色,自动排版打印版
很多系统限制 20–200 KB
做一张 900×383 头图 + 500×500 次图,分享到朋友圈不裁坏。
标题放中间三分之一:朋友圈只显示中间 1:1
或直接用社交尺寸导出的「公众号头图」预设
字号 ≥ 60px 在手机上才清楚
同一张图输出 900×383 与 500×500
发给客户的样片、发到网上的作品统一打上署名。
文字或 Logo,选平铺或角落,透明度 30% 左右
照片里的 GPS 位置与设备信息发布前建议清除
把几张聊天 / 界面截图处理成可以公开发的一张图。
头像、姓名、手机号;马赛克比模糊更不易还原
人多的合照用这个省事
截图加圆角看起来更像一张卡片
把一张图切成 9 张按顺序发。
九宫格只显示正方形
也可切 3×1、2×3
扫描的老照片去划痕、上色、放大。
先纠偏与去阴影
本地模型修复与上色
最后调色与锐化
一张 Logo 生成 favicon、Apple Touch Icon、PWA 图标。
主体留 15% 边距,圆形遮罩下不切角
输出 16 / 32 / 180 / 192 / 512 与 manifest 片段
微信单个文件 25 MB 内(会话)或朋友圈 5 分钟内。
分辩率、码率、时长决定怎么压
时长少一半体积少一半
选「目标大小 25 MB」或 720p / 1.5 Mbps
从视频里截几秒做成聊天用 GIF。
去黑边、聚焦主体
宽 320–480、10–15 fps 体积可控
微信表情 ≤ 1 MB
讲座 / 采访视频只要声音。
选 MP3 或无损 AAC 直接复制流
环境噪声大时
多段音频音量不一致时统一到 -16 LUFS
手机拍的几段合成一条。
横竖混拼会出现黑边
参数一致时无损拼接,否则重新编码
段间过渡不生硬
下载的 SRT 与视频对不上,或播放器只认另一种格式。
SRT / VTT / ASS 互转,整体提前或延后 N 秒
GBK 字幕在 Mac / 手机上乱码,转 UTF-8
录口播 / 语音备忘并去噪、去静音。
拿到一段响应,搞清楚状态码、JSON 结构、token 内容。
4xx 是请求问题,5xx 是服务端问题,看卡片里的处理建议
报错定位到行列;去转义处理被二次编码的字符串
看 exp 是否过期、payload 里的字段
响应里的 Unix 时间戳换成可读时间
线上与本地的配置文件差在哪。
YAML 与 JSON 混着时先转成同一种
缩进与键顺序一致后差异才有意义
并排高亮
服务器每天凌晨备份,确认表达式没写错。
输入后看未来 10 次执行时间对不对
「进程与服务」组的 crontab 条目有完整示例
755 让 cron 能执行
一条很长的 docker run 转成 compose.yaml 好维护。
未识别的参数会明确提示
systemctl enable docker、usermod -aG docker
把 10.0.0.0/16 拆给几个部门。
输入 CIDR 看主机范围,按需拆分
ip addr、ip route、ss、nc
本地写好的代码变成一个远程仓库。
init → add → commit → remote add → push -u 五步
gitignore.io、Choose a License 在「Git 与协作」组
先在模拟器里推一遍再上真机
把链接 / Wi-Fi / 名片做成二维码放进海报。
Wi-Fi 二维码扫码即连;容错率选 H 以便加 Logo
打印前用手机或本工具识别确认
印刷用 SVG 或 ≥ 1000px PNG
交付日倒计时、请假天数、生日农历。
两个日期差几天 / 加减 N 天
排除周末与法定假日
把 12345.6 写成「壹万贰仟叁佰肆拾伍元陆角」,算个税。
注册新账号时生成密码,并在本地查看 TOTP 验证码。
16 位以上含符号
密钥不落盘,作为验证器 App 的临时替代
英寸转厘米、磅转公斤、外币换人民币。
长度 / 重量 / 温度 / 面积
需要联网取汇率