discord-transcriptify

Save any Discord channel as a single HTML file that looks just like Discord.

Live demo · Quick start · What it renders · Options · FAQ

Close a support ticket, archive a channel before deleting it, or keep a record of a moderation case: one function call gives you a file that anyone can open in a browser, with the same colours, message grouping, embeds, buttons and Components V2 as the Discord client. Built for discord.js v14 and v15.

Why discord-transcriptify

• Looks like Discord. Markdown, mentions in role colours, replies, reactions, embeds, polls, voice messages, forwarded messages, threads, stickers and system messages.
• Full Components V2. Containers, sections, media galleries, separators, files, checkboxes and radio groups, next to the classic buttons and select menus.
• One file, nothing to host. Search, member list, pinned messages, image preview and a thread panel are built in. Images and voice messages can be embedded so the file keeps working after Discord's links expire.
• Safe with untrusted content. Everything is escaped, and the image downloader blocks private addresses, redirects and oversized files. See Security.
• Easy on small bots. Memory limits are on by default, and sharp is optional.
• Typed and tested. Written in TypeScript, tested on Node 18, 20, 22 and 24, and published from GitHub Actions with npm provenance.

Quick start

npm install discord-transcriptify
# or: pnpm add discord-transcriptify
# or: yarn add discord-transcriptify

const { createTranscript } = require('discord-transcriptify');

const attachment = await createTranscript(channel);
await channel.send({ files: [attachment] });

With TypeScript or ES modules:

import { createTranscript } from 'discord-transcriptify';

Example: transcript when a ticket closes

const { createTranscript } = require('discord-transcriptify');

client.on('interactionCreate', async (interaction) => {
  if (!interaction.isButton() || interaction.customId !== 'close-ticket') return;
  await interaction.deferReply();

  const attachment = await createTranscript(interaction.channel, {
    saveImages: true, // keep the images even after the channel is gone
    filename: `${interaction.channel.name}.html`,
  });

  const logs = await client.channels.fetch(LOG_CHANNEL_ID);
  await logs.send({ content: `Ticket closed by ${interaction.user}`, files: [attachment] });
  await interaction.channel.delete();
});

Want to see the result first? Open the live demo, a transcript of a made-up support ticket. It is generated by demo/generate.ts on every push.

What it renders

• Messages: Discord markdown (bold, quotes, code blocks with syntax highlighting or ANSI colours, spoilers, timestamps), mentions in their role colour, replies, edits, reactions, slash command headers, and consecutive messages from the same author grouped under one header.
• Embeds: every field, including author, footer, thumbnail and image.
• Components V2: container (accent colour, spoiler), text display, section with a thumbnail or button accessory, media gallery, file, separator, checkbox and radio group.
• Buttons and select menus: all styles, link buttons, emoji and the disabled state. A select menu opens on a click and lists its options, with their emoji and descriptions.
• Attachments: images, videos and audio inline; several images as one grid; any other file as a card with its name, size and a download link.
• Voice messages: waveform, duration and a play button.
• Polls: answers, vote bars and the poll result message. The closing time is shown as a date, never as time left.
• Forwarded messages: content, attachments, embeds and components of the forwarded snapshot.
• Threads: the "started a thread" message with its preview, and optionally every message inside the thread.
• Stickers: PNG, APNG and GIF as images. Lottie stickers show a placeholder with the sticker name, since they need a player script.
• System messages: joins, pins, boosts and thread creation.
• Server tags and role icons next to author names.

Built-in interface

The file also carries a small interface for whoever reads it. All of it can be switched off.

Message search	Member list
Pinned messages	Thread panel

Clicking an image opens it full-screen.

Requirements

• Node.js 18.17 or newer
• discord.js 14.19 or newer, or v15. Components V2 arrived in 14.19.
• sharp is optional. Install it only if you want embedded images resized or compressed.

Usage

From a channel

createTranscript fetches the messages for you.

const { createTranscript } = require('discord-transcriptify');

const attachment = await createTranscript(channel, {
  limit: -1, // every message in the channel
  saveImages: true, // embed images so the file does not depend on Discord's CDN
});

await channel.send({ files: [attachment] });

From messages you already have

const { generateFromMessages } = require('discord-transcriptify');

const attachment = await generateFromMessages(messages, channel);

messages is an array or a discord.js Collection of messages.

Getting a string or a buffer

const { createTranscript, ExportReturnType } = require('discord-transcriptify');

const html = await createTranscript(channel, { returnType: ExportReturnType.String });

returnType accepts 'attachment' (the default), 'buffer' and 'string'.

Options

Every option is optional. All of them work with both functions, except limit and filter, which belong to createTranscript.

Option	Default	What it does
limit	-1	How many messages to fetch. -1 fetches the whole channel.
filter	none	(message) => boolean. Messages that return false are left out, thread messages included.
returnType	'attachment'	'attachment', 'buffer' or 'string'.
filename	'transcript-{channel name}.html'	File name of the attachment.
saveImages	false	Download images and embed them in the file. See Embedding images.
saveVoiceMessages	false	Download voice messages and embed them. Accepts { maxSizeKb, maxTotalSizeKb }. See Memory use.
saveThreadMessages	false	Fetch the messages of each thread and show them under its preview. Accepts { limit }, default 500 per thread.
fetchMembers	true	Ask Discord for the members behind the authors, so names show in their role colour. See Role colours.
ui	true	The interactive interface. false removes it; an object switches single features. See Interactive UI.
hydrate	false	true renders the components on the server. With false the page loads the component scripts from a CDN when it is opened.
footerText	'Exported {number} message{s}.'	Text of the footer.
poweredBy	true	Show the "Powered by discord-transcriptify" line in the footer.
favicon	'guild'	'guild' uses the server icon; any other string is used as the icon URL.
callbacks	built in	Your own resolveChannel, resolveUser, resolveRole, resolveImageSrc and resolveVoiceSrc.

Threads

saveThreadMessages covers both kinds of thread: one started from a message and one created on its own, which the channel shows as a "started a thread" line. Each thread costs one extra API request per 100 messages. Up to three threads are fetched at a time. When a thread is longer than the limit, the most recent messages are kept and the transcript says how many were left out. A limit below 1 means no limit.

Role colours

Messages fetched from a channel arrive without their member, and the role colour of an author lives on the member. With fetchMembers on, the uncached authors are requested from the gateway, 100 per request. If a request fails, the transcript is still generated and those names stay uncoloured. Pass fetchMembers: false to send no request at all.

Interactive UI

By default the file includes a small amount of JavaScript. Each feature has its own switch:

await createTranscript(channel, {
  ui: {
    imagePreview: true, // click an image to see it full-screen
    search: true, // search the messages
    memberList: true, // members and how many messages each one sent
    pinnedMessages: true, // list of pinned messages
    voicePlayer: true, // play button and waveform on voice messages
    threadModal: true, // open a thread in a panel beside the channel, and list the saved threads in the header
    serverTags: true, // server tag next to the author's name
    selectMenus: true, // close an open select menu on a click elsewhere or Escape
  },
});

ui: false removes all of it. The transcript stays complete without scripts: voice messages fall back to the browser's audio controls and thread messages to a collapsible block. Select menus still open and close on a click.

The scripts are self-contained: none is loaded from another site and none sends data anywhere. The server tag badge is an image served by Discord, like an avatar. Server tags do not appear with hydrate: true, because a server-rendered page carries no profile data for the script to read.

Embedding images

With saveImages: true, images are downloaded and written into the file as data URLs. This covers image attachments, embed images, thumbnails, author and footer icons, and Components V2 media galleries and thumbnails. By default:

• images are resized to fit 800×800 and compressed at 75% quality (JPEG and WebP; PNG stays lossless), if sharp is installed; without it they are embedded untouched
• files over 10 MB are skipped and keep their original URL
• once the embedded images add up to 50 MB, the remaining ones keep their original URL
• four images are downloaded at a time
• GIF and SVG files are embedded as they are
• videos are not embedded

Pass an object to change the limits:

saveImages: {
  maxSizeKb: 5120, // per image
  maxTotalSizeKb: 20480, // all images together
  concurrency: 2, // downloads at the same time
  quality: 60,
  maxWidth: 1200,
  maxHeight: 1200,
  webp: true, // convert to WebP, usually smaller
  sharpCache: false, // turn off sharp's cache of decoded images
},

For full control, give resolveImageSrc your own function. It is called once per image and returns a string to use as the image source, undefined to keep the original URL, or null to leave the image out of the transcript.

The built-in downloader can also be configured and passed in:

const { createTranscript, TranscriptImageDownloader } = require('discord-transcriptify');

await createTranscript(channel, {
  callbacks: {
    resolveImageSrc: new TranscriptImageDownloader()
      .withMaxSize(5120) // in KB, checked against the bytes actually received
      .withMaxTotalSize(20480) // in KB, all images together
      .withConcurrency(2) // downloads at the same time
      .withMaxDimensions(1200, 1200) // resize to fit, needs sharp
      .withCompression(60, true) // 60% quality, convert to webp, needs sharp
      .withTimeout(10_000, 30_000) // headers and body timeouts, in ms
      .withSharpCache(false) // see Memory use
      .build(),
  },
});

If neither .withCompression() nor .withMaxDimensions() is called, sharp is never loaded. The total set with .withMaxTotalSize() belongs to the function .build() returns, so build one per transcript.

Memory use

Everything a transcript embeds is held in memory until the file is written, so a channel with many images can use more memory than a small bot has. Three limits keep that bounded, and all of them apply by default when saveImages is on:

Limit	Option	Default
Size of one image	saveImages.maxSizeKb	10240 (10 MB)
Size of all images together	saveImages.maxTotalSizeKb	51200 (50 MB)
Downloads running at once	saveImages.concurrency	4

Voice messages have their own pair: saveVoiceMessages.maxSizeKb (default 25600) and saveVoiceMessages.maxTotalSizeKb (default 51200).

The totals are counted on what ends up in the file, after compression. When a total is reached, the transcript is still generated: the remaining images or voice messages keep their Discord URL instead of being embedded. Zero or less turns a total, or the concurrency limit, off.

On a bot with little memory, lower the totals and install sharp, which shrinks each image before it is embedded:

await createTranscript(channel, {
  saveImages: { maxTotalSizeKb: 10240, maxSizeKb: 2048, concurrency: 2, sharpCache: false },
});

sharpCache: false turns off sharp's cache of decoded images, which holds up to 50 MB. The setting applies to the whole process. Set it here rather than with sharp.cache(false) in your bot: the copy of sharp the library imports is loaded on the first image and turns the cache back on.

The file itself is also built in memory, so expect a peak of a few times the size of the finished transcript. If that is still too much, do not embed at all: give resolveImageSrc a function that uploads the image to your own storage and returns its URL.

Security

A transcript is built from content that anyone in the channel can write, so the library treats all of it as untrusted.

Markup is escaped. Message content, names and every other value from Discord are rendered through React, so HTML in a message is shown as text and never executed.

Downloads are guarded. A Components V2 gallery, a thumbnail or an embed can point at any URL, not only at Discord's CDN. When saveImages or saveVoiceMessages is on, the built-in downloader refuses:

• any scheme other than http: and https:
• loopback, private, link-local and reserved addresses, including the cloud metadata address 169.254.169.254. This applies to addresses written literally and to hostnames that resolve to one, and it is checked when the connection is made, so DNS rebinding does not get past it
• redirects
• responses over the size limit, measured on the bytes received and not on the size the message claims

If your images really live on an internal host, opt out with new TranscriptImageDownloader().withAllowPrivateNetwork(). Only the address rule is relaxed. Do this only if you trust everyone who can post in the channel.

A resolveImageSrc or resolveVoiceSrc function of your own gets none of these checks: validating the URLs it receives is up to you.

To report a vulnerability, see SECURITY.md.

FAQ

Messages show up empty. Your bot needs the Message Content intent, both in the Developer Portal and in the client: intents: [GatewayIntentBits.Guilds, GatewayIntentBits.GuildMessages, GatewayIntentBits.MessageContent].

Images in old transcripts stopped loading. Discord's attachment links expire. Pass saveImages: true to embed the images in the file when it is created.

Names are not coloured. The colour comes from the author's member, which has to be fetched. Keep fetchMembers on; see Role colours.

The file is too big to send. Discord limits uploads by server boost level. Lower saveImages.maxTotalSizeKb, install sharp, or use returnType: 'buffer' and store the file somewhere else.

Does it work without a bot? The library renders discord.js messages, so it needs a client to fetch them. To render messages you already have, use generateFromMessages.

Development

npm install
npm run build
npm test
npm run demo   # writes the demo transcript to demo/out

npm test needs no bot and no network. tests/README.md describes the rest, and CONTRIBUTING.md covers pull requests.

Credits

discord-transcriptify started from discord-html-transcripts, used under the Apache-2.0 license.

License

Apache-2.0. See LICENSE and NOTICE.