The file type is detected by checking the magic number of the buffer.
This package is for detecting binary-based file formats, not text-based formats like .txt, .csv, .svg, etc.
We accept contributions for commonly used modern file formats, not historical or obscure ones. Open an issue first for discussion.
Install
npm install file-type
This package is an ESM package. Your project needs to be ESM too. Read more. For TypeScript + CommonJS, see . If you use it with Webpack, you need the latest Webpack version and ensure you configure it correctly for ESM.
[!IMPORTANT]
File type detection is based on binary signatures (magic numbers) and is a best-effort hint. It does not guarantee the file is actually of that type or that the file is valid/not malformed.
Robustness against malformed input is best-effort. When processing untrusted files on a server, enforce a reasonable file size limit and use a worker thread with a timeout (e.g., make-asynchronous). These are not considered security issues in this package.
Usage
Node.js
Determine file type from a file:
import {fileTypeFromFile} from 'file-type';console.log(await fileTypeFromFile('Unicorn.png'));//=> {ext: 'png', mime: 'image/png'}
Determine file type from a Uint8Array/ArrayBuffer, which may be a portion of the beginning of a file:
import {fileTypeFromBuffer} from 'file-type';import {readChunk} from 'read-chunk';const buffer = await readChunk('Unicorn.png', {length: 4100});console.log(await fileTypeFromBuffer(buffer));//=> {ext: 'png', mime: 'image/png'}
An example is @tokenizer/http, which requests data using HTTP-range-requests. A difference with a conventional stream and the tokenizer, is that it can ignore (seek, fast-forward) in the stream. For example, you may only need and read the first 6 bytes, and the last 128 bytes, which may be an advantage in case reading the entire file would take longer.
import {S3Client} from '@aws-sdk/client-s3';import {makeChunkedTokenizerFromS3} from '@tokenizer/s3';import {fileTypeFromTokenizer} from 'file-type';// Initialize S3 clientconst s3 = new S3Client();// Initialize the S3 tokenizer.const s3Tokenizer = await makeChunkedTokenizerFromS3(s3, { Bucket: 'affectlab', Key: '1min_35sec.mp4'});// Figure out what kind of file it is.const fileType = await fileTypeFromTokenizer(s3Tokenizer);console.log(fileType);
Note that only the minimum amount of data required to determine the file type is read (okay, just a bit extra to prevent too many fragmented reads).
Returns a Promise which resolves to the original readable stream argument, but with an added fileType property, which is an object like the one returned from fileTypeFromFile().
This method can be handy to put in a stream pipeline, but it comes with a price.
Internally stream() builds up a buffer of sampleSize bytes, used as a sample, to determine the file type.
The sample size impacts the file detection resolution.
A smaller sample size will result in lower probability of the best file type detection.
Supports the following options in addition to the general options:
sampleSize
Type: number
Default: 4100
The sample size in bytes.
Example
import {fileTypeStream} from 'file-type';const url = 'https://upload.wikimedia.org/wikipedia/en/a/a9/Example.jpg';const response = await fetch(url);const stream = await fileTypeStream(response.body, {sampleSize: 1024});if (stream.fileType?.mime === 'image/jpeg') { // stream can be used to stream the JPEG image (from the very beginning of the stream)}
supportedExtensions
Returns a Set<string> of file extensions that can be detected.
supportedMimeTypes
Returns a Set<string> of MIME types that can be detected.
Options
customDetectors
Array of custom file type detectors to run before default detectors.
For example:
import {fileTypeFromFile} from 'file-type';import {detectXml} from '@file-type/xml';const fileType = await fileTypeFromFile('sample.kml', {customDetectors: [detectXml]});console.log(fileType);
maxBufferSize
Type: number
Default: Infinity
Maximum size in bytes of an individual buffer allocated to read file content, or a stream lookahead request.
Must be a positive safe integer or Infinity.
File content can determine how much data a detector requests, for example through a field specifying the length of a metadata entry.
This opt-in limit bounds the memory allocations needed to fulfill those reads, including reads made by custom detectors.
Setting a finite cap can help prevent excessive read-buffer allocations driven by untrusted file content (CWE-789: Memory Allocation with Excessive Size Value).
Built-in reads already have format-specific bounds; choose a finite cap to suit your application.
An oversized request rejects detection with strtok3.BufferSizeError, which exposes requestedSize and maxBufferSize.
During gzip content probing, this error is suppressed and detection falls back to gz; the cap still prevents the oversized allocation.
This is not a total memory limit.
It does not cover buffers allocated directly by detectors or fileTypeStream() sampling.
When using fileTypeFromTokenizer(), configure maxBufferSize on the input reader itself; this option does not override that reader’s limits.
The built-in format-specific limits, including ZIP text-entry limits, apply independently of this option.
For FileTypeParser.toDetectionStream(), configure detection options such as maxBufferSize through the parser constructor.
The method only accepts stream options such as sampleSize.
mpegOffsetTolerance
Default: 0
Specifies the byte tolerance for locating the first MPEG audio frame (e.g. .mp1, .mp2, .mp3, .aac).
Allows detection to handle slight sync offsets between the expected and actual frame start. Common in malformed or incorrectly muxed files, which, while technically invalid, do occur in the wild.
A tolerance of 10 bytes covers most cases.
Custom detectors
Custom file type detectors are plugins designed to extend the default detection capabilities.
They allow support for uncommon file types, non-binary formats, or customized detection behavior.
Detectors can be added via the constructor options or by directly modifying FileTypeParser#detectors.
Detectors provided through the constructor are executed before the default ones.
[!NOTE]
Not safe for concurrent use. Create a new instance per call if you need concurrency.
Example adding a detector
import {FileTypeParser} from 'file-type';import {detectXml} from '@file-type/xml';const parser = new FileTypeParser({customDetectors: [detectXml]});const fileType = await parser.fromFile('sample.kml');console.log(fileType);
Available third-party file-type detectors
@file-type/av: Improves detection of audio and video file formats, with accurate differentiation between the two
@file-type/cfbf: Detects Compound File Binary Format (CFBF) based formats, such as Office 97–2003 documents and .msi.
@file-type/pdf: Detects PDF based file types, such as Adobe Illustrator
@file-type/xml: Detects common XML file types, such as GLM, KML, MusicXML, RSS, SVG, and XHTML
Detector execution flow
If a detector returns undefined, the following rules apply:
No Tokenizer Interaction: If the detector does not modify the tokenizer's position, the next detector in the sequence is executed.
Tokenizer Interaction: If the detector modifies the tokenizer's position (tokenizer.position is advanced), no further detectors are executed. In this case, the file type remains undefined, as subsequent detectors cannot evaluate the content. This is an exceptional scenario, as it prevents any other detectors from determining the file type.
Writing your own custom detector
Below is an example of a custom detector. This can be passed to the FileTypeParser via the customDetectors option.
import {FileTypeParser} from 'file-type';const unicornDetector = { id: 'unicorn', // May be used to recognize the detector in the detector list async detect(tokenizer) { const unicornHeader = [85, 78, 73, 67, 79, 82, 78]; // "UNICORN" in ASCII decimal const buffer = new Uint8Array(unicornHeader.length); await tokenizer.peekBuffer(buffer, {length: unicornHeader.length, mayBeLess: true}); if (unicornHeader.every((value, index) => value === buffer[index])) { return {ext: 'unicorn', mime: 'application/unicorn'}; } return undefined; }}const buffer = new Uint8Array([85, 78, 73, 67, 79, 82, 78]);const parser = new FileTypeParser({customDetectors: [unicornDetector]});const fileType = await parser.fromBuffer(buffer);console.log(fileType); // {ext: 'unicorn', mime: 'application/unicorn'}
/**@param tokenizer - The [tokenizer](https://github.com/Borewit/strtok3#tokenizer) used to read file content.@param fileType - The file type detected by standard or previous custom detectors, or `undefined` if no match is found.@returns The detected file type, or `undefined` if no match is found.*/export type Detector = { id: string; detect: (tokenizer: ITokenizer, fileType?: FileTypeResult) => Promise<FileTypeResult | undefined>;};
Abort signal
Some async operations can be aborted by passing an AbortSignal to the FileTypeParser constructor.
import {FileTypeParser} from 'file-type';const abortController = new AbortController();const parser = new FileTypeParser({signal: abortController.signal});const promise = parser.fromStream(blob.stream());abortController.abort(); // Abort file-type reading from the Blob stream.
Supported file types
MIME media subtypes prefixed with x-ft- are custom and defined by us. They are neither formally registered with IANA nor based on any informal conventions.
3g2 - Multimedia container format defined by the 3GPP2 for 3G CDMA2000 multimedia services
3gp - Multimedia container format defined by the Third Generation Partnership Project (3GPP) for 3G UMTS multimedia services
ISO 9660 requires random-access input or fileTypeStream() with sampleSize >= 32_774; fileTypeFromStream() is unsupported.
When [Content_Types].xml cannot be reached — it may trail the content entries, and only so much of what precedes it is scanned — the type is guessed from the archive's directory names: word/, xl/ and ppt/ are reported as docx, xlsx and pptx. Macro-enabled and template variants are therefore reported as their base type, and a non-OOXML archive containing one of those directories is reported as OOXML.
Pull requests are welcome for additional commonly used file types.
The following file types will not be accepted, but most of them are supported by third-party detectors.