Video to Text: Transcribe Any Video in 99 Languages
Upload the video — MP4, MOV, WEBM, MKV, AVI, up to 5 GB — and a modern AI video transcriber returns roughly 92–97% accurate text on clear speech, with speaker labels and timestamps, in a few minutes. No audio-extraction step, no format conversion, no manual timing. The output can be a plain transcript, a DOCX for editing, or an SRT/VTT subtitle file — from one processing pass.
By VexaScribe Editorial · Updated
Transcribe a Video in 4 Steps
Works the same for MP4 screen recordings, MOV iPhone captures, WEBM YouTube exports, or MKV archives.
Drop the video into the uploader
Drag your MP4, MOV, WEBM, MKV, AVI, or WMV file directly. No pre-conversion, no ffmpeg step, no audio extraction — modern transcription engines read the audio track out of the container natively. Files up to 5 GB (~5-6 hours of standard 720p video).
Let language detection do its job
99 languages, auto-detected from the first speech in the audio track. Only override if the video opens with music or long silence. Language identification runs before transcription — one file, one detection pass.
Review speakers and check names
Speaker diarization labels up to 50 voices (best accuracy with 2-6 speakers). AI gets proper nouns wrong more than anything else — spot-check names, companies, and numbers against the video timeline.
Export in whichever format you actually need
TXT or DOCX for a transcript document; SRT or VTT for subtitles; the same processed file gives you both, no re-upload. Or use AI chat over the transcript: ask "what did they decide about pricing?" and jump to the timestamped moment.
Cost reference: a 1-hour video costs about $0.30 on the Starter plan ($2/mo, 200 minutes) — full comparison against pay-per-file and human video transcription services on the cost calculator.
Transcript vs Subtitles — Which Do You Actually Need?
Same processing pass, two different output formats. The choice depends on what you're making with the text:
Plain transcript (TXT / DOCX) — this page
- Reading, editing, or quoting the spoken content
- Show notes, meeting minutes, research analysis
- Blog posts, articles, book excerpts from interviews
- Sending to a writer/editor for further work
Subtitles (SRT / VTT) — different pages
- Overlaying captions on video for YouTube/TikTok/Reels
- Accessibility compliance (closed captions)
- Multilingual subtitle tracks for global distribution
- Burned-in social-media-style animated captions
For pure SRT output, see SRT Generator. For a listicle comparing subtitle tools with styling and burn-in, see Best Subtitle Generation Tools.
You can get both from one upload. The same processed video gives you TXT/DOCX and SRT/VTT — no re-upload. Pick this page if the transcript document is your primary output; pick the subtitle pages if the subtitle file is what you're after.
Video to Text vs Video Transcription vs Video Transcriber
Different searches, same job. If you landed here from any of these, you're in the right place — the tool below produces every output the phrasing implies.
| Search term | What it actually asks for |
|---|---|
| Video to text | The generic search — expects a converter that produces readable text output. That's what this page is. |
| Video transcription | The industry term for the same job. Emphasizes the workflow (transcribe = write down what's said) more than the file format. |
| Video transcriber | The tool doing the work. "AI video transcriber" specifically implies an automated engine, not a human service. |
| Video transcript | The output — the written document. "Video transcript generator" and "video transcript maker" ask for the same thing. |
| Convert video to text | Same as "video to text" — the "convert" phrasing carries a slight expectation of format options (TXT vs DOCX vs SRT). |
Supported Video Formats
Rule of thumb: if the file plays in VLC or QuickTime, it'll transcribe. The container matters less than the audio track inside it.
| Format | What it usually means | Support |
|---|---|---|
| MP4 | H.264 or H.265 — the default for phones, screen recorders, and most editors | Universal |
| MOV | Apple QuickTime — iPhone recordings, Mac screen captures, ProRes exports | Universal |
| WEBM | Browser-native recordings, YouTube downloads, OBS captures | Universal |
| MKV | High-quality archives, ripped DVDs, hi-fi lecture recordings | Universal |
| AVI | Older Windows recordings; still common in enterprise | Universal |
| WMV | Legacy Windows Media — old training decks and archives | Works |
| FLV | Legacy Flash video, still around in old CMSs | Works |
| MPEG/MPG | Older camcorders, satellite captures | Works |
| ProRes RAW / editing codecs | Uncompressed intermediates from Final Cut / DaVinci — no audio track sometimes | Export to MP4 first |
The 5 GB per-file cap covers roughly 5-6 hours of standard 720p video or 2-3 hours of 1080p. Larger files: compress to 720p with Handbrake (free) or split with ffmpeg. Video resolution has no effect on transcript accuracy — the audio track is what the engine reads.
What the Video Transcript Actually Looks Like
Speaker labels, second-precision timestamps, natural paragraph breaks at speaker turns. The video transcript generator formats the output for readability by default; SRT export uses the same timestamps in subtitle format.
[00:00:00] SPEAKER 1: All right, we're live. Thanks for making time — I know it's early on your side. [00:00:04] SPEAKER 2: No problem. Where do you want to start? [00:00:07] SPEAKER 1: Let's do the Q3 numbers first, then the platform stuff. Ada shared a doc last night — did you see the churn figures? [00:00:15] SPEAKER 2: Yeah, the enterprise cohort is at 4.2% monthly. Still elevated but down from 5.1 in June. [00:00:23] SPEAKER 1: That's the one. What's driving the drop?
The generated transcript is editable in-browser before export — fix a proper noun once, it stays fixed. SRT export takes the same speaker turns and slices them at subtitle-appropriate boundaries.
Free Video to Text — What 30 Minutes Gets You
Honest version first: 30 minutes free covers most single-video jobs. A typical 45-minute lecture uses 30 min of your balance and stops; a 20-minute meeting fits comfortably. The balance doesn't reset monthly — it's a trial, not a recurring quota. If you need ongoing free use, the honest options are:
- OpenAI Whisper installed locally — free forever, unlimited, offline. Requires Python + GPU for reasonable speed. Best for privacy-sensitive video.
- YouTube auto-captions — free if you can upload the video as Unlisted. Accuracy is significantly lower (~60-70% on accented English) than modern Whisper-based tools.
- Free web tools with daily caps — typically 1-3 files/day, 25 MB per file (about 1-2 minutes of HD video). Fine for short clips, useless for meetings.
Translate Video to Text (Non-English Sources)
Two different jobs get called "translate video to text" — make sure you're asking for the right one:
Transcribe in the source language
Spanish video → Spanish transcript. Whisper Large-v3 auto-detects the language and produces text in the same language. 99 supported languages, no configuration needed.
Translate to English (or 132 other languages)
Spanish video → English transcript, Japanese video → French transcript, and so on. Handled via a two-stage flow: Whisper Large-v3 transcribes in the source language, then a translation stage outputs the target. For the dedicated video-input translation flow with platform workflows (YouTube, TikTok, Instagram, Zoom) and timestamp preservation, use translate video to text. For the audio-only flow, use translate audio to text.
Cross-language transcription is a two-stage pipeline: transcribe first (source language), translate second (any of 133 targets). Standard tier uses Google Translate for the translation stage; Premium tier uses an internal translation model for higher accuracy on client-facing work.
Language-pair dedicated pages: Spanish · French · German · Chinese · Japanese · Russian.
How to Transcribe a Video (Manual Alternatives)
If you're evaluating AI transcription against the alternatives, the honest comparison:
Manual transcription (typing while listening)
Roughly 4–6 hours per hour of video for a careful transcript with speaker labels. Accurate but slow. Reasonable only for very short clips or when audio is genuinely too damaged for AI (heavy overlap, phone-bandwidth crosstalk, single-mic multi-speaker crowd audio).
Human transcription service (Rev, GoTranscript)
$1–$1.50/min ($60–$90 per hour of video), 12–24 hour turnaround, 99% accuracy on clean audio. Right for legal/medical/insurance work where the accuracy delta from AI matters. Overkill for internal meetings, YouTube videos, or lectures.
AI transcription (this page's workflow)
About $0.30 per hour of video on $2/mo plans, 5–10 minute turnaround, 92–97% accuracy on clean audio. Right for meetings, lectures, podcasts, content creation, research interviews, and any use case where a 3–8% word error rate is acceptable (usually with a quick pass of manual editing).
Common Video Transcription Use Cases
The workflow is identical; the export format is what changes. Six of the most common paths:
How to Create a Transcript from a Video (5 Methods)
Five ways to generate a transcript from a video file, ranked by how much effort and money each requires. The right method depends on how much video you have, how privacy-sensitive it is, and whether you need publication-quality output.
Method 1 — Upload to an AI Web Tool (Fastest)
Drop the MP4/MOV/MKV into VexaScribe (this page), HappyScribe, Descript, or Rev. Processing takes 5-10 minutes per hour of video. Cost: US$0.05-0.60 per hour of video. Best for volume, speed, and standard business use.
Steps: (1) create an account (30 min free trial), (2) drag the video into the uploader, (3) wait for processing, (4) review speaker labels and proper nouns, (5) export as TXT/DOCX/SRT/VTT.
Method 2 — Whisper Installed Locally (Free, Requires Setup)
Install OpenAI Whisper on your computer (Python + pip install openai-whisper). Run it against any video file. Cost: US$0 forever, unlimited. Setup time: 30-60 minutes if you've never used Python. Requires a GPU for acceptable speed (roughly real-time on RTX 3060, 10× real-time on RTX 4090).
Best for high-volume, privacy-sensitive, or budget-constrained work. See how accurate Whisper is for real WER benchmarks.
Method 3 — Native OS Transcription (Microsoft Word, Google Docs)
Microsoft Word Transcribe: in Microsoft 365 Word (web version), Home tab → Dictate dropdown → Transcribe → Upload audio. Free with M365 subscription (comes free with .edu email); 300 minutes/month cap. Extract audio from video first via FFmpeg or an online converter.
Google Docs Voice Typing: Tools → Voice typing. Real-time only (no file upload), Chrome only, English-focused. Play the video with your speakers and let Google Docs transcribe the audio via your microphone. Free unlimited, but accuracy suffers from the double-mic-path.
Method 4 — Hire a Human Transcription Service (Most Accurate)
Rev Human ($1.50-1.99/min), TranscribeMe Verbatim ($1.25/min), GoTranscript ($0.90/min). 12-48 hour turnaround. 99%+ accuracy. Cost for a 1-hour video: US$54-120. Best for legal, medical, publication-quality work where every word must be verifiable.
For the honest hybrid workflow that gets 99% at 1/5 the cost: use Method 1 for a rough draft, then hire a human to verify only the segments you'll actually quote.
Method 5 — Manual Typing (Playback + Type)
Play the video in a media player (VLC, QuickTime) and type into a document. Use a foot pedal (Infinity IN-USB-2 ~$70) to control playback hands-free. Realistic pace: 4-6 hours of work per 1 hour of audio. Only justified for very short clips (under 5 minutes) or when audio quality is so poor that AI fails.
Which method should you use? Method 1 (AI web upload) covers 90% of use cases. Method 2 (Whisper local) if you need free/unlimited/private. Method 4 (human) for publication-critical work. Method 3 (native OS) only if you already have M365 or Google Docs and low-volume needs. Method 5 (manual) only for very short clips.
Video Transcript Generator vs Manual Creation
A video transcript generator is any tool that produces text from video automatically — VexaScribe, HappyScribe, Descript, Rev, and Whisper installed locally all qualify. The generator does three things: (1) extracts the audio from the video container (MP4/MOV/MKV), (2) runs the audio through a speech recognition model (Whisper Large-v3 or comparable), (3) formats the output as TXT/DOCX/SRT/VTT with speaker labels and timestamps.
Manual creation means playing the video and typing what you hear — the pre-AI standard. Realistic time cost: 4-6 hours of typing per 1 hour of video for a proficient typist. Realistic monetary cost if you value your time at $30/hr: $120-180 per hour of video. Compared to AI generator cost of $0.05-0.60 per hour of video, the generator is 200-3,000× cheaper.
The only case where manual creation still makes sense: your video is 30 seconds long, you don't have internet, and you're fluent in the speaker's language. Otherwise, use a generator (Method 1 or 2 above).
Get Transcript from Video — Quick Method for MP4/MOV Files
The 60-second workflow for getting a transcript from a video file (no signup required for the trial): drag the MP4/MOV/MKV into the uploader above → click “Transcribe” → the transcript appears in the editor once processing completes. Copy to clipboard or export as TXT/DOCX/SRT. That's it.
If your video is stored in a cloud service (Google Drive, Dropbox, OneDrive), copy the shareable link and paste it into VexaScribe — we'll fetch the file for you. If it's a YouTube video, paste the YouTube URL into our YouTube transcription flow — no need to download the file first.
For MP4 specifically: the transcript generator extracts the audio track from the MP4 container automatically. You don't need to extract audio to MP3 first — skip that step. See our MP4 to text guide for MP4-specific details.
When to Reach for a Different Tool
You only have a YouTube URL, not a file — use YouTube Transcription. Paste the URL directly — no download step. YouTube's own captions are used when available, otherwise the audio is re-transcribed by Whisper.
TikTok, Instagram Reel, or short social clip — use Platform-specific tools. The social tools handle URL input for those platforms directly. Faster than downloading the file first.
You only need the SRT file, not the transcript document — use Video to SRT. Shorter workflow, same engine — skips the transcript editor entirely and hands you the timed .srt.
You want a summary of the video, not a full transcript — use Video Summarizer. The summarizer runs the same transcription then generates structured key points, decisions, and action items on top.
The audio is the source — no video track — use Audio to Text. For M4A voice memos, MP3 podcasts, or WAV field recordings the format-specific guides give a shorter path.
Have the video ready?
Drop the file — speaker labels, timestamps, 99 languages, TXT/DOCX/SRT/VTT export. 30 minutes free, no credit card, plans from $2/mo.
Transcribe Video FreeFrequently Asked Questions
How do I transcribe a video to text?
Four steps: (1) drop the video file into an AI transcription tool (VexaScribe, Otter, Rev, Trint — modern tools accept MP4, MOV, WEBM, MKV, AVI directly without pre-extracting audio); (2) let it auto-detect the language (99 languages supported on Whisper-based tools); (3) spot-check speaker labels and proper nouns against the video; (4) export as TXT, DOCX, SRT, or VTT depending on whether you need a transcript document or subtitle file. Processing takes 5-10 minutes per hour of video. No conversion to MP3 or WAV needed — that's advice from 2018-era tools.
What's the best free video transcription tool in 2026?
Depends on use case. For one-off videos: VexaScribe gives 30 minutes free on signup, accepts up to 5 GB per file (most free tools cap at 25 MB, about 1-2 minutes of HD video). For unlimited free with privacy: OpenAI Whisper installed locally on your computer — free forever, offline, requires Python setup. For a YouTube video you own: YouTube auto-captions are free but ~60-70% accuracy vs 92-97% on modern Whisper-based tools. For ongoing free web use: TurboScribe offers 3 free files/day (30-min cap each). Avoid unknown "free video to text" sites — they often use pre-2022 speech engines with significantly worse output.
How does a video transcript generator work?
A video transcript generator has three parts: (1) audio extraction — pulling the audio track out of the video container (MP4, MOV, WEBM) without a manual ffmpeg step; (2) automatic speech recognition — modern generators use Whisper Large-v3 or purpose-built engines like Deepgram Nova-3 to convert speech to text with 92-97% accuracy on clean audio; (3) formatting — speaker diarization, timestamps, and export to your chosen format (TXT for reading, DOCX for editing, SRT for YouTube subtitles, VTT for HTML5 video). The whole pipeline runs in the cloud — you don't install anything.
Can I get a transcript AND subtitles from the same video?
Yes — and you should, in one upload. The same transcription pass produces the timestamped text; you then export it once as TXT/DOCX (transcript for blog posts, research, show notes) and once as SRT/VTT (subtitle file for YouTube, TikTok, Reels, Premiere Pro, Final Cut, DaVinci Resolve). The timestamps are identical — they're just serialized differently. Don't upload the video twice.
Do I need to extract audio from the video first?
No. All modern AI transcription tools (VexaScribe, Whisper, AssemblyAI, Deepgram, Rev AI) accept video files directly and extract the audio track internally. The "extract audio first" advice comes from older online converters (2018-era Google Cloud Speech v1, IBM Watson) that only accepted audio inputs. With current tools, drag your MP4 / MOV / WEBM / MKV / AVI in directly.
What video formats can I transcribe?
Universal formats that always work: MP4 (H.264 or H.265), MOV (Apple QuickTime — iPhone recordings, Mac screen captures), WEBM (browser recordings, YouTube exports), MKV (high-quality archives), AVI (older Windows recordings), and WMV. Also works: FLV, MPEG/MPG. Not directly supported: ProRes RAW and proprietary editing codecs — export those to MP4 first from your editor. Rule of thumb: if the file plays in VLC or QuickTime, it transcribes. Video resolution has no effect on transcript accuracy — the audio track is what the engine reads.
How long can the video be? What's the file size limit?
VexaScribe accepts up to 5 GB per file — roughly 5-6 hours of standard 720p video or 2-3 hours of 1080p HD. Most free online tools cap at 25 MB (only about 1-2 minutes of HD video). Otter and Notta free tiers fail on any HD video longer than ~5-10 minutes. For files larger than 5 GB: compress to 720p with Handbrake (free) or split with ffmpeg. Reducing video bitrate aggressively won't hurt transcript accuracy — only audio quality matters.
How accurate is an AI video transcriber?
Modern Whisper-based transcribers (VexaScribe, AssemblyAI, Deepgram) achieve 92-97% word accuracy on clear English audio per the Open ASR Leaderboard and OpenAI's published Whisper benchmarks. Accuracy is governed by AUDIO quality, not video quality — a 4K video with bad mic audio transcribes worse than a 480p video with good mic audio. Accuracy drops on: heavy accents (85-92%), background music or noise (80-90%), technical jargon (medical, legal, engineering), and overlapping speakers. Tools that claim "99.9% accuracy" are using marketing language — the actual peer-reviewed WER (word error rate) for state-of-the-art ASR is 3-8% on clean speech.
Can I translate a video to text in English?
Yes, if the source is non-English. Whisper's translation mode takes a video in Spanish, French, German, Mandarin, etc. and outputs an English transcript. The output is idiomatic English (not word-for-word translation) and works best when source audio is clear. Cross-language transcription in the other direction — Spanish audio to French transcript, for example — requires two passes: transcribe in the source language, then translate the text with a separate tool. Whisper's built-in translation only targets English.
How is video transcription different from YouTube auto-captions?
Two very different things. YouTube's auto-captions use Google's older ASR engine — accuracy is roughly 60-70% on accented English and drops fast for non-English content. Fine for low-stakes viewing. Modern AI video transcription (VexaScribe, Whisper, AssemblyAI) uses transformer-based models trained on hundreds of thousands of hours of audio — 92-97% accuracy on clean English, with speaker labels, professional formatting, and export to TXT/DOCX/SRT/VTT. For publishing, business, or research use, the accuracy gap is dramatic enough that YouTube auto-captions aren't a real substitute.