How to Transcribe an MP4 Video to Text | Free, No Upload, SRT Subtitles Included

Created on:November 21, 2026 at 10:37 AMReading time: 16 min
thumbnail Image

When you want the words in a recorded video or a meeting recording as text, a search turns up many options: Transcribe in Microsoft Word, ChatGPT, YouTube's automatic captions and a long list of cloud transcription services. All of them can turn speech into text, but they differ in whether they accept a video file as it is, whether the file is sent to an outside server, and whether the result can be saved as a subtitle file, and each free plan has its own cap.

This article first compares those methods using their official help and pricing pages, and then walks through transcribing an MP4 video in the browser without uploading it, from getting the file ready to exporting text or subtitles.

What You'll Learn from This Article 

  • How the ways to transcribe a video differ
  • How to transcribe an MP4 video step by step
  • How to prepare each kind of video
  • Which model to use for an English video
  • Which export format to choose for text and for subtitles

All you need is Chrome or Edge on a PC and the video file you want to transcribe. There is no account, no app install, no video upload and no conversion to an audio file at any point.

Ways to transcribe a video, compared 

The ways to transcribe a video fall into three groups: features built into a word processor, an AI chat or a video site; cloud transcription services; and tools that do the processing inside the browser. The table below summarizes what each provider's official help and pricing pages said when I checked them on November 21, 2026.

Method

Sent to a server

What you can use for free

Reads an MP4 video

SRT subtitle export

Transcribe in Microsoft Word

Yes (sent to Microsoft and saved to OneDrive)

A Microsoft 365 subscription feature, up to 300 minutes of uploaded audio per month

Yes (.wav, .mp4, .m4a and .mp3)

Not stated on the official page

Voice typing in Google Docs

Not stated on the official page

Not stated on the official page

No file import is described (it types what the microphone hears)

Not stated on the official page

ChatGPT

Yes (file upload)

Audio uploads are not available on the Free plan

MP4 files identified as video are not supported (audio-only MP4 is)

Not stated on the official page

Gemini

Yes (file upload)

Up to 5 minutes of video in total (1 hour with Google AI Pro or Ultra)

Videos can be uploaded, up to 2 GB each (formats are not listed)

Not stated on the official page

YouTube automatic captions

Yes (the video is published to YouTube)

Not stated on the official page

Only videos uploaded to YouTube

Not stated on the official page

Cloud transcription services

Yes (file upload)

Monthly minutes are capped on the free plans (second table below)

Formats are not listed on the pricing pages

Varies by service (second table below)

A tool that runs in the browser (Transcriber)

No (only processing statistics are sent)

No limit on length or on the number of runs

Yes (MP4, MOV, MKV, WebM and more)

SRT and VTT

The sources are Microsoft's "Transcribe your recordings", Google's "Type & edit with your voice", OpenAI's "Uploading files and audio to ChatGPT", Google's "Upload & analyze files in Gemini Apps" and YouTube's "Use automatic captioning". "Not stated on the official page" means the page does not mention it, not that the feature is missing. YouTube's page also notes that automatic captions can misrepresent what was said because of mispronunciations, accents, dialects or background noise.

Cloud transcription services cap how much you can use on their free plans. Here is what the pricing pages of three widely used services said in November 2026.

Service

Free plan allowance

Length per recording

SRT export

"Otter.ai" (Basic)

300 minutes per month, with 3 file imports per user in total

Up to 30 minutes

Pro plan and above ($16.99 per month on monthly billing); Basic exports MP3 and TXT

"Rev" (Free)

45 minutes of AI transcription and captions per month

Not stated on the pricing page

Not stated on the pricing page

"Descript" (Free)

60 media minutes per month

Not stated on the pricing page

Not stated on the pricing page (video exports are watermarked and limited to 720p)

The tables only reflect what the official pages say, and I have not used Word, ChatGPT, Gemini or these cloud services to transcribe video myself. To turn video into text I use "Transcriber", a transcription tool I built that runs in the browser. What sets it apart from the other methods in the table is that the video file never leaves the device and that there is no cap on length or on the number of runs.

I transcribe video for three things: subtitles for videos I post to social media and YouTube, turning meeting and seminar recordings into text, and keeping the content of X videos I have saved as text. I use the same tool for all three, so the steps in this article use Transcriber as well.

What you need 

  • The video file you want to transcribe (MP4, MOV, MKV, WebM, AVI, WMV, FLV, MPG, MPEG-TS or 3GP)
  • Chrome or Edge on a PC (processing is faster where WebGPU is available)
  • A network connection the first time, to download the model (about 500 MB for Parakeet Ultra, the English default on a PC with integrated graphics)

Step 1: Get the video file ready 

The video needs to be a file saved on your PC. How you get it there depends on the format and on where the video came from, so read the case that matches yours.

Videos shot on a phone 

Videos from the iPhone camera are saved with the ".mov" extension, and they load without being converted to MP4. You can also pick the video directly in a phone's browser, but a phone can process a shorter length at once than a PC, so move long videos to a PC first.

Meeting and webinar recordings 

A recording of an online meeting loads as it is, as long as it is saved as a video file such as MP4. For a recording with several people talking, leave "Speaker diarization" on (the default), and each line is labeled with who said it, as "Speaker A", "Speaker B" and so on. The steps for transcribing Zoom and Teams recordings are on Transcriber's "Transcribe a meeting recording into minutes" page.

Videos you posted on YouTube 

Videos on YouTube and other video sites cannot be loaded from their URL. For a video you uploaded yourself, save it as an MP4 first and then load that file (the steps for saving it from YouTube Studio are on Transcriber's "Transcribe an MP4 video" page).

Videos on X 

To transcribe a video from X, save it with "X Video Downloader", which I also built, and press "Transcribe" in its list to hand the saved MP4 straight to Transcriber. Transcriber opens in a new tab, receives the video file from X Video Downloader's server and starts transcribing with whatever settings are in place at that moment. X Video Downloader deletes the video file on its server 24 hours after processing finishes, so the transcription has to be done before then.

How to Save Videos from X (Formerly Twitter) | Step-by-Step for PC and Phone | Ryusei.IO

Save videos posted on X (formerly Twitter), and archives of live broadcasts, as MP4 without installing an app or creating an account. Covers the steps on both PC and phone, and why some videos cannot be saved.

faviconryusei.io

Videos edited in video editing software 

For a video that will get subtitles, load the finished version you exported after cutting and editing. Subtitle timing counts from the start of the loaded video, so subtitles made from the raw footage drift from the finished video by however much was cut.

Step 2: Choose the model and language, then load the video 

Once the video is ready, open Transcriber in the browser.

Transcribe MP4 Video — Free, No Upload, Subtitle-Ready | Transcriber

Transcribe MP4 video inside your browser for free. The video is never uploaded — only its audio track is processed on your device. Check and fix captions while the video plays. MOV, MKV, WebM and AVI are supported, with SRT subtitle export.

favicontranscriber.tools.ryusei.io
  1. Open Transcriber's "Transcribe an MP4 video" page.
  2. Under "Language", choose the language spoken in the video. For an English video, leave it on "English".
  3. Check "Model". If you do not choose one, a default suited to the language and the device is selected (see "Choosing a model" below).
  4. "Speaker diarization" is on by default. Turn it off for a video with only one speaker, or when you are making subtitles.
  5. Drop the video file onto the area marked "Drop audio or video files, or click to choose (multiple files OK)", or click it to choose the file. There is no start button, and transcription begins the moment the file is loaded, so settle steps 2 to 4 first.
The settings for the model (Parakeet Ultra), the language (English) and speaker diarization, and the area where the video file is dropped

Once transcription starts, the progress and the recognized lines appear one after another. The first time a model is used, it has to be downloaded before processing begins. Closing the tab stops the processing, so keep it open until it finishes.

The screen partway through transcribing a video, with the progress and the recognized lines appearing in order

Choosing a model 

The model is chosen automatically from the language and the device's capabilities. For English, the default is "Parakeet Ultra" on a PC with integrated graphics in Chrome or Edge, "Whisper Large v3 Turbo" on a PC with a dedicated GPU ("Whisper Small" on older GPUs), "Whisper Small" on a Mac with Apple silicon, and "SenseVoice Small" on phones, on PCs without WebGPU and in Safari or Firefox on PCs with integrated graphics. Other models can be picked yourself from the "Model" list.

The model list opened, with Parakeet Ultra selected and SenseVoice Small, Whisper Large v3 Turbo and other models listed

For this article I joined 63 English utterances from a public read-speech dataset (FLEURS) into an MP4 of about 10 minutes (606.2 seconds) and compared three models in Chrome on a PC with a Ryzen 7 8745H and its integrated Radeon 780M graphics, the same hardware as my own PC. The processing times are from the second run, after the models had been downloaded. The error rate is the word error rate (WER): the exported text is compared with the reference transcript word by word, after differences in capitalization, punctuation and the way numbers are written have been normalized away.

Model

Runs on

Processing time

Speed relative to real time

Word error rate

Number of subtitles

Parakeet Ultra (default)

GPU (WebGPU)

32.8 s

18.5x

5.4%

125

SenseVoice Small

CPU (WASM)

77.4 s

7.8x

9.9%

81

Whisper Large v3 Turbo

GPU (WebGPU)

127.8 s

4.7x

6.0%

67

The processing times are with speaker diarization on; with it off, Parakeet Ultra took 24.7 seconds and the text was the same. The default, Parakeet Ultra, had both the lowest error rate and the shortest processing time, and Whisper Large v3 Turbo came close on error rate but took nearly four times as long. The test video is read speech with the volume of each utterance evened out, scored on a single sample of 1,318 words, so the numbers will differ for conversations or videos with background music.

The videos I transcribe for my own use are in Japanese, and for those I keep the Japanese default, Haas-ja v1.0. The English figures above therefore come only from the test video made for this article. The number of subtitles differs by model because each model decides where to break lines in its own way; how long the resulting subtitles are is covered in "How to Create an SRT File".

Step 3: Play the video with captions and fix the result 

When transcription finishes, the video can be played in the player on the result screen. The line at the current playback position is shown as a caption over the lower part of the video, so you can compare the wording with the picture while you check for mistakes.

The player on the result screen, with the transcribed line shown as a caption over the lower part of the video

Captions are switched on and off with the captions button, the second icon from the right below the video. To fix a mistake, hover over the line in the list and click the pencil icon that appears; for a line with the wrong timing, click the line's time and retype the start and end. Clicking a line moves playback to that line. The corrections carry over to the exported file as they are.

When it outputs words nobody said 

Transcription models sometimes output text that is not in the audio, which is called "hallucination". On the English test video none of the three models produced such a line, but the gaps between its utterances are only 0.7 seconds and every utterance was brought to the same volume. When I measured a Japanese video that contained barely audible utterances, Whisper Large v3 Turbo alone output a Japanese phrase meaning "Thank you for watching" in two places where nobody had said it.

Lines like these tend to appear in silent stretches and in stretches with only background music, so play the video with captions on and check that no text shows up where nobody is speaking. To remove such a line, choose "Delete line" from the line's "⋮" menu.

Fixing the same misheard word in one pass 

Names and technical terms are often misheard the same way more than once. With "Find & replace" on the result screen, every occurrence of the same mistake can be fixed in one operation instead of line by line.

Step 4: Export as text or as subtitles 

Once the corrections are done, choose a format from "Download" below the result. There are seven formats: TXT, Word and PDF under "Documents", SRT and VTT under "Subtitles", and CSV and JSON under "Data", and every one of them is generated inside the browser.

The "Download" menu, split into Documents (TXT, Word, PDF), Subtitles (SRT, VTT) and Data (CSV, JSON)

For the three things I transcribe video for, the formats that fit are these.

Purpose

Format that fits

Why

Turning meeting and seminar recordings into text

Word (DOCX) or TXT

The text can be edited into minutes before sharing

Subtitles for videos posted to social media and YouTube

SRT (VTT for video players on web pages)

Each line carries its start and end times

Keeping the content of saved X videos

TXT

It can be pasted into a notes app and searched

If you only want to paste the text into a notes app, "Copy text" puts it on the clipboard without making a file.

For subtitles 

When exporting an SRT file for subtitles, turn "Speaker diarization" off in step 2. If it stays on, every subtitle starts with a speaker name such as "Speaker A: ", and that name is shown on screen as part of the subtitle.

Line length depends on how the model breaks lines. In my measurement, the default Parakeet Ultra produced subtitles with a median of 59 characters and a maximum of 146, and each subtitle was a single line with no line break inside it. Netflix's "English (USA) Timed Text Style Guide" allows 42 characters per line and two lines at most, and 25 of the 125 subtitles were longer than those 84 characters. There is no feature that splits subtitles automatically at a fixed length, so split long lines with "Split at cursor" on the result screen before exporting. What goes into an SRT file, and how to split lines and fix timing, are covered in detail in "How to Create an SRT File".

To draw the subtitles into the picture so that they show in any player, load the exported SRT into the "Burn subtitles into video" page of "Media Tools", a web app I built. The steps are in "How to Add Subtitles to a Video in Your Browser".

Translating a video in another language into English 

For a video in another language, choose that language under "Language" in step 2 before loading the file. When transcription finishes, open "Translate", choose English as the language to translate into and press "Translate": the translation appears under each line, and choosing "Both" from the captions button plays the original and the translation over the video in two rows.

A Japanese video with the Japanese original and the English translation shown over it in two rows

Translation uses the translator built into Chrome 138 or later and Edge 148 or later on a PC, so in other browsers and on phones the "Translate" panel only shows a notice that it is not available. The translation is also generated on the device, so the transcript is never sent anywhere to be translated. The first time a language is used, its translation model is downloaded. In my test, translating a 3-minute Japanese video (25 lines) into English took about 7 seconds including that download, and about 1 second once the model was in place.

The translation is machine translation, and in my test one line lost a word: a Japanese sentence saying that many Parisians are self-centered, arrogant and rude came out as "It is said that many Parisians are self-centered and rude.", without "arrogant". With "Both" chosen from the captions button, the original and the translation are shown together, so a line like this can be compared with the original while the video plays. I use translated subtitles myself, playing the translated SRT in VLC and on Transcriber's result screen. When exporting, the "Original", "Translation" and "Both" switch to the left of "Copy text" decides which text goes into the file.

Videos that transcribe poorly 

Depending on the video, accuracy can drop or transcription can stop even when every step above is followed. Here are two causes that are easy to identify.

Loud background music or noise 

When background music or ambient noise is loud, the voice is recognized less accurately. For a video with background music, one option is to make a copy without the music first, with the "Remove Music from Video" page of Media Tools, and load that copy into Transcriber. The page removes only the music, and speech, narration and sound effects remain. Like Transcriber, it processes the video in the browser, and it runs in Chrome and Edge on a PC where the GPU (WebGPU) is available.

Videos with no audio 

A screen recording made with the microphone and system audio turned off in the recording settings contains no voice at all. Transcriber processes only the video's audio, so if it stops with an error, play the video first and check that a voice can be heard. Other fixes are collected under "If something goes wrong" on Transcriber's "Transcribe an MP4 video" page.

Frequently asked questions 

Is it really free? Are there limits on length or on the number of runs?

It is free, and there is no limit on length or on the number of runs. The transcription runs on the visitor's own device, and my server does not process it.

Do I need an account or a login?

You do not. Opening the page and loading a video is all it takes, and no email address is asked for.

Is there any chance the video file is sent to a server?

There is not. Loading the video, extracting the audio and transcribing it all happen inside the browser, so the video file never leaves the device. Apart from downloading the models, Transcriber sends statistics used to fix problems (such as device specs and error types) to the site's own server, but they do not include the audio, the transcript, file names or IP addresses. You can turn this off with the switch in the "Help us improve the service" section near the bottom of the page.

Can I transcribe a video in Microsoft Word?

You can, with a Microsoft 365 subscription. According to Microsoft's help page, Transcribe accepts .mp4 files, sends the audio to Microsoft and allows up to 300 minutes of uploaded audio per month, and the page says nothing about exporting subtitles.

Can ChatGPT transcribe a video?

Not as a video file. OpenAI's help page says MP4 and WebM files identified as video are not supported by audio uploads, and audio uploads are not available on the Free plan.

Can I transcribe a video on an iPhone alone?

You can. In a phone's browser, the default model for English is "SenseVoice Small". The length that can be processed at once is shorter than on a PC, so for long videos see "Videos shot on a phone" in step 1.

Can I transcribe a YouTube video from its URL?

You cannot. Transcriber reads only video files saved on the device. For a video you posted yourself, save it as an MP4 first and then load it.

Summary 

Of the methods in the comparison table, Transcriber is the one that reads an MP4 file as it is, keeps it on your device and puts no cap on length or on the number of runs. Word, ChatGPT, Gemini and the cloud services all need the file uploaded, and each of them either caps the minutes or leaves the feature out of its free plan. Transcribing an MP4 video with Transcriber takes these four steps.

  1. Save the video as a file on your PC
  2. Set "Language", "Model" and "Speaker diarization", and then load the video file
  3. Play the video with captions on the result screen and fix misheard words and lines nobody said
  4. Export from "Download" as Word or TXT for text, or as SRT for subtitles

For a video that will get subtitles, turn "Speaker diarization" off in step 2 before loading it. If the SRT is exported with the default setting left on, every subtitle starts with a speaker name such as "Speaker A: ", and that name also appears over the video.

Related articles and tools 

How to Add Subtitles to a Video in Your Browser | Free, No Install, No Upload | Ryusei.IO

No install, no upload. Transcribe a video, fix the SRT, and burn the subtitles into the picture — all inside your browser, free and without an account.

faviconryusei.io
How to Transcribe an m4a File | No Conversion, No Upload | Ryusei.IO

A step-by-step guide to transcribing m4a recordings from the iPhone Voice Memos app without converting the format. No upload and no account are needed, and separating speakers and exporting subtitles are done entirely in the browser.

faviconryusei.io
Remove Music from Video — Keep Voice and Sound Effects (Free, No Upload) | Media Tools

Free background music remover for video. AI strips the music and keeps speech and sound effects, and the picture is copied without re-encoding. Runs in your browser, so the video is never uploaded. No sign-up. For Chrome and Edge on PC.

faviconmedia.tools.ryusei.io

Latest Tips