Document conversion tools that run on your machine, not ours
PDF, Word, Excel, Markdown, HTML, images, audio and video. Fifteen converters, and the interesting question with all of them is what survives the trip.
16 tools here, 150+ across the site
DOCX→
PDF→
MD→
HTML
each hop keeps less than the last
All tools
Audio CompressorShrink audio file size with adjustable bitrate presetsFile & Document
Audio ConverterConvert audio between MP3, WAV, AAC, OGG and FLACFile & Document
EPUB to PDF ConverterConvert EPUB ebooks to PDF or extract PDF text into a reflowable EPUBFile & Document
Excel to CSV ConverterConvert XLSX sheets to plain CSV formatFile & Document
Excel to PDF ConverterConvert XLSX sheets to a PDF fileFile & Document
HTML to PDF ConverterConvert any HTML content to a PDF fileFile & Document
Image ConverterConvert any image from any formatFile & Document
Markdown to PDF ConverterRender Markdown documents as styled PDFsFile & Document
MP3 to WAV ConverterConvert compressed MP3 audio to lossless WAV formatFile & Document
MP4 to GIF ConverterConvert video clips to animated GIFs for sharingFile & Document
PDF to Excel ConverterExtract tables from PDFs into Excel spreadsheetsFile & Document
PDF to Word ConverterConvert PDF files to editable Word documentsFile & Document
Video CompressorCompress MP4, MOV and WebM videos up to 90% smallerFile & Document
Video to Audio ExtractorStrip audio track from any video fileFile & Document
WAV to MP3 ConverterCompress WAV audio files to smaller MP3 formatFile & Document
Word to PDF ConverterConvert DOCX files to PDF with formatting preservedFile & Document
Fidelity
What actually survives each conversion
Every format stores a different set of things. Converting means mapping one set onto another, and whatever has no equivalent is dropped silently.
| Conversion | Text | Layout | Images | What you lose |
|---|---|---|---|---|
| Word to PDF | Kept | Kept | Kept | Editability, and any tracked changes or comments |
| PDF to Word | Kept | Approximated | Kept | Exact positioning. PDF stores glyphs at coordinates, not paragraphs. |
| Excel to CSV | Kept | Lost | Lost | Formulas, formatting, multiple sheets, charts. CSV holds one grid of text. |
| Markdown to PDF | Kept | Kept | Kept | Nothing meaningful. Markdown is simple enough to map cleanly. |
| HTML to PDF | Kept | Mostly | Kept | Interactivity, and anything positioned by JavaScript after load |
| Video to audio | n/a | n/a | Lost | The entire visual track, which is the point |
PDF to Word is the hard one, and it is worth understanding why. A PDF does not contain paragraphs or tables. It contains instructions to draw specific glyphs at specific coordinates. Converting back means inferring structure that was thrown away when the PDF was created. Any converter promising perfect fidelity is overselling. Simple single column documents come back well; complex multi column layouts with floating figures do not.
A scanned PDF is a different problem again. It contains photographs of pages, not text at all, so extraction returns nothing until optical character recognition has run.
Under the hood
The libraries doing the actual work
These are not wrappers around a conversion API. They are real libraries running in your browser, which is why the first load of the heavier ones takes a moment.
ffmpeg.wasmThe full ffmpeg toolchain compiled to WebAssembly. Handles every audio and video conversion, compression and extraction on this page. Around 25MB on first load, then cached.
SheetJSReads and writes xlsx, xls, csv and several older spreadsheet formats. Handles the awkward parts such as merged cells and date serial numbers.
html2canvasRasterises a DOM subtree to a canvas, which is how HTML becomes an image or a PDF page.
jsPDFBuilds PDF documents byte by byte in the browser, including page breaks, fonts and embedded images.
Canvas 2DThe native browser API behind every image conversion, resize and recompression step.
Web Audio APIDecodes compressed audio into raw samples, used by the WAV converter with no external library at all.
The tradeoff is honest. A server with a real CPU converts a large video faster than your laptop will. What you get instead is that the video never left your laptop, there is no queue, no file size ceiling and no retention policy to read.
Gotchas
Six things that surprise people mid conversion
Small details that account for most of the confusion in this category.
| Situation | What is happening |
|---|---|
| CSV opens with mangled dates | Excel guesses at date formats on import. A column of 03/04 values becomes March or April depending on your regional settings, not the file. |
| Leading zeros vanish | Postcodes, phone numbers and product codes get read as numbers. The fix is importing as text rather than opening directly. |
| Semicolons instead of commas | Many European locales use semicolons because the comma is their decimal separator. Both are valid CSV. |
| Extracted MP3 sounds worse | Extracting audio from an already compressed video and re-encoding it compounds two lossy passes. Copying the stream without re-encoding avoids this where the format allows. |
| Word to PDF loses fonts | If a font is not embedded, the renderer substitutes a similar one and line breaks shift. Embedding fonts at export time prevents it. |
| PDF text will not copy | Either the PDF is a scan, or its font encoding has no mapping back to Unicode. Both need OCR rather than extraction. |
Browse more
The wider toolbox
Every category on ConvertNow, free and growing.
From the blog
Format guides and gotchas
What each format stores, and what quietly disappears in conversion.
FAQ
Questions about document conversion tools
Fidelity, size limits and the conversions that lose things.
Are my files uploaded to a server?
No. Conversion runs through libraries loaded into the page, including ffmpeg compiled to WebAssembly for media and SheetJS for spreadsheets. The file is read from your disk into browser memory and never transmitted. There is no temporary server copy and nothing to request deletion of.
Is there a maximum file size?
No hard cap, because there is no upload. Your device memory is the practical limit. A large video takes longer rather than being rejected, and you are not sharing a processing queue with anyone else.
Why does PDF to Word lose my formatting?
Because a PDF does not contain paragraphs or tables. It contains instructions to draw specific glyphs at specific coordinates. Converting back means inferring structure that was discarded when the PDF was made. Simple single column documents recover well, complex multi column layouts do not.
Why does nothing extract from my PDF?
It is almost certainly a scan, meaning the pages are photographs rather than text. Extraction finds nothing because there is no text layer to find. That needs optical character recognition first, which the image to text tool handles.
Why do my CSV dates look wrong when I open them in Excel?
Excel guesses at date formats on import using your regional settings, not the file. A column of 03/04 values becomes March or April depending on where your copy thinks it is. Importing as text rather than double clicking the file avoids the guess entirely.
Where did my leading zeros go?
Excel read the column as numbers. Postcodes, phone numbers, product codes and account numbers all lose leading zeros this way. The fix is the same, import as text rather than opening the CSV directly, or prefix the values before export.
Why is the first video conversion so slow?
The ffmpeg WebAssembly build is around 25MB and downloads on first use, then caches. After that, conversion speed depends on your CPU. A server with dedicated hardware would be faster, but the file would have to leave your machine to get there.
Which formats are actually supported?
Documents cover PDF, Word, Excel, CSV, Markdown and HTML in various directions. Media covers MP4, MP3, WAV, GIF and the common audio and video containers. Images cover JPEG, PNG, WebP and AVIF. Each converter lists its own accepted inputs.
Processing
Explore all tools
Conversion without the upload
Documents are the most sensitive thing people convert. Contracts, invoices, medical letters, financial statements.
Browser side librariesConversion runs through libraries loaded into the page, such as SheetJS for spreadsheets and ffmpeg compiled to WebAssembly for media. The file never leaves your device.
Larger files are slower, not blockedBecause your machine does the work, there is no upload size cap. A big video simply takes longer rather than being rejected.
Nothing to delete afterwardsThere is no temporary file on a server, no expiry window and no deletion request to submit. Closing the tab is the deletion.