How to Transcribe a Video to Text for Free, Without Uploading It

To transcribe a video to text for free, run OpenAI's Whisper on your own computer. It reads the video file directly, understands about 100 languages, has no minute cap, and exports plain text or subtitle files. You can drive it from a command line, from a free app such as Buzz or Subtitle Edit, or from a desktop app like Sablate.

The online "free video to text" sites work too, but they ration it: three files a day, thirty minutes each, an account, an upload. Below are the three local ways to do it, what each costs you in setup, and how to get a transcript that does not need an hour of cleanup.

Why free online transcribers stop at a limit

Every minute of speech recognition costs someone GPU time. When you upload a video to a transcription website, that someone is the website, so the free tier is a sample with a meter on it.

The meters look similar across the category. TurboScribe, one of the most generous, gives free users three transcriptions a day, up to 30 minutes per file, one upload at a time; longer files and unlimited use sit behind its Unlimited plan, around $10 a month billed yearly. Other services cap the free tier at a few minutes of audio, watermark the export, or hold back SRT and DOCX until you pay. None of this is unfair. It is the arithmetic of renting out GPUs.

Run the same model on your own machine and the arithmetic flips. Your computer pays for the compute, so nobody needs to meter it. The trade is that you supply the hardware and the patience: a laptop without a dedicated GPU transcribes slowly, and the first run downloads a model of anything from about 78 MB to 3 GB.

The engine is often identical. TurboScribe says it is built on Whisper, as do many of its competitors. A free local transcription and a paid cloud one can come out of the same model, and the real difference is who owns the GPU and where the file goes.

The three local ways to transcribe a video for free

1. The Whisper command line. Free, unlimited, every model. Needs Python and FFmpeg, and gives you files, not an editor.
2. A free Whisper app: Buzz or Subtitle Edit. Free, unlimited, a window instead of a terminal. Light on editing.
3. Sablate. A desktop app built around correcting the transcript and turning it into subtitles in several languages. Free with limits, $29 one-time for the full version.

All three keep the video on your disk.

Method 1: Whisper from the command line

The reference implementation is OpenAI's openai-whisper package. It needs Python 3 and FFmpeg, which Whisper uses to pull the audio out of the video file.

Install FFmpeg first. On Windows:

winget install Gyan.FFmpeg

On macOS, with Homebrew:

brew install ffmpeg

Then install Whisper and transcribe:

pip install -U openai-whisper
whisper "interview.mp4" --model turbo --language en --output_format txt

That writes interview.txt next to where you ran the command. Swap txt for srt or vtt to keep the timings, or use all to get every format at once. --model turbo is large-v3-turbo; on a machine without a GPU, --model small finishes in a fraction of the time.

The limitations are real:

• On Windows, pip installs a CPU build of PyTorch. Using an NVIDIA GPU means reinstalling PyTorch from the CUDA index on pytorch.org, which is where most people's first evening goes.
• There is no editor. You get a text file. Correcting it means opening it in a text editor and listening to the video in another window.
• Silence and music can produce invented text. Whisper sometimes repeats a line or writes "Thanks for watching" over a quiet stretch. Our guide to Whisper hallucinations covers why and which settings reduce it.

If you are comfortable in a terminal, this is the most flexible free option there is, and it scales to a folder of files with a one-line loop.

Method 2: A free Whisper app (Buzz or Subtitle Edit)

If you would rather click than type, two free apps wrap the same models in a window.

Buzz is an open-source (MIT) transcription app for Windows, macOS and Linux. Import a video, pick a model and a language, press Run, and export TXT, SRT or VTT. It can use several engines, including faster-whisper and whisper.cpp, supports GPU acceleration, and can label speakers. The Windows installer is not code-signed, so Windows shows a SmartScreen warning that you have to click past.

Subtitle Edit is the long-standing free subtitle editor for Windows, with a macOS build in beta. Its Video menu has an Audio to text (Whisper) option that downloads the engine and model for you, transcribes, and opens the result in a full timing editor. It reads and writes more subtitle formats than any other tool, and you can save the result as plain text when you only want the words.

Both are excellent and both are free with no limits. The cost is mostly in the second half of the job: Buzz gives you a transcript viewer with playback and search but little for subtitle timing, and Subtitle Edit gives you a powerful editor that assumes you already know what a cue, a frame rate and a subtitle format are.

Method 3: A desktop app like Sablate

Sablate runs faster-whisper locally on Windows and Mac and is built around what happens after the transcription: fixing it, and turning it into subtitles.

1. Drop the video file into the app and choose the spoken language and a model tier: Fast, Balanced, Accurate (large-v3-turbo) or Best (large-v3).
2. The transcript opens in an editor where every line is tied to its moment in the video. Click a line to hear it, then correct it in place.
3. Add names and product terms to the glossary before you transcribe, and use find and replace (Ctrl+H) to fix a misheard word everywhere at once.
4. Export .txt or copy the text, or export .srt and .vtt. If the video needs subtitles in other languages, add translation layers to the same transcript instead of transcribing again.

The free tier is permanent and needs no account, but it is the most limited of the local options here: videos up to 10 minutes, three jobs a month, and the Fast and Balanced models. Your first video longer than that runs at full length once. If you only ever need one long transcript, Buzz or the command line will do it for nothing, and that is the honest recommendation.

Where Sablate earns its $29 one-time Pro price is volume and the editor: unlimited length, every Whisper model, batch transcription, and one-click export of every language a video has.

Free video transcription tools compared

Prices and limits as published in October 2026. The pattern holds across the category: free local tools have no meter, online tools need no installation, and that convenience is the thing you pay for in minutes.

Is free Whisper as accurate as paid transcription?

On clear speech, the accuracy of a free local transcription is the accuracy of the model you choose, not of the price you paid. A paid service running large-v3 and your laptop running large-v3 produce near-identical text from the same audio.

The model choice matters far more than the tool. large-v3-turbo is large-v3 with its decoder cut from 32 layers to 4: about eight times faster, with a difference in the finished transcript that is small on ordinary footage. tiny and small are fast but drop words on accents, crosstalk and fast speech. Our Whisper model comparison has the sizes, the speeds and a documented case where the bigger model skipped speech a smaller one caught.

Three things no engine fixes for you, free or paid:

1. Proper names and jargon. Whisper spells an unfamiliar surname phonetically, and differently each time. A glossary helps most at the start of a file, where the model reads it as context; a find and replace pass catches the rest.
2. Silence and music. Long quiet stretches are where hallucinated lines appear. Trim dead air or check those spots first.
3. Overlapping speakers. Whisper writes one stream of text and does not label who said what. Buzz adds a speaker identification step and paid services offer speaker labels; Sablate and the Whisper command line do not.

Paid services still win on two things: they run on a phone, and some add speaker labels and summaries on top. If you need neither, paying is optional.

Which method should you use?

"I need one transcript, once, and I'm comfortable in a terminal." Use the Whisper command line with --model turbo if you have a GPU, --model small if you do not.

"I want a free app and I'll never touch a terminal." Install Buzz. It is free with no limits and exports the formats you need.

"I actually need subtitles, and the timing has to be right." Use Subtitle Edit if you already know subtitle editing, or Sablate if you want the transcript, corrections and subtitles in one place. Our guide to making an SRT file covers the format rules either way.

"I'm on my phone, or the video is short and not sensitive." An online tool's free tier is the fastest path. Three files a day is plenty for occasional use.

"The video is already on YouTube and I only need to read it." YouTube's own Show transcript panel is free and instant for videos with captions. No tool required.

"I transcribe every week, or the video then needs subtitles in other languages." That is the case Sablate is built for: unlimited length in Pro, a real editor, and translation layers on the same transcript.

How to get a cleaner transcript the first time

Most cleanup comes from four avoidable mistakes.

Set the language explicitly. Auto-detection listens to the first 30 seconds. A video that opens with music or a greeting in another language gets transcribed in the wrong one, or comes out partly translated into English, a known Whisper quirk.

Use the biggest model your machine runs comfortably. Moving from small to large-v3-turbo removes more errors than any amount of post-editing, and on a GPU it costs little time.

Fix names before anything else. Correct the transcript before you export subtitles or translate it. A name fixed once at the source is fixed in every file you make from it.

Keep TXT and SRT apart in your head. TXT is for reading; SRT is for timing. If you might need subtitles later, export SRT now as well, since rebuilding timings from a plain text file is the slowest job in this whole workflow.

Start with one video and the tool that fits your situation above. If you want the transcript, the corrections and the subtitles in one window, download Sablate for Windows or Mac; the free tier needs no account.

Frequently asked questions

Can I transcribe a video to text completely for free?

Yes. OpenAI's Whisper is free, open source and runs on an ordinary Windows or Mac computer, so there is no per-minute charge and no daily cap. Online services that are free usually ration it instead: TurboScribe's free plan, for example, allows three files a day of up to 30 minutes each. Locally, the only limit is how long you are willing to wait.

Is free Whisper accurate enough to replace paid transcription?

For clear speech, usually yes. Several paid transcription services run Whisper themselves, so the price rarely changes the engine. What changes accuracy is the model size: large-v3-turbo and large-v3 are far better than tiny or small on accents, noise and fast speech. Names and jargon need correcting with any engine, free or paid.

Do I need a GPU to transcribe video with Whisper?

No. Whisper runs on the CPU, only more slowly. An NVIDIA GPU is typically five to ten times faster than the same machine on CPU, which is what makes the large models practical for long recordings. Without a GPU, the small model is the realistic everyday choice and turbo is something you leave running.

Can I transcribe a video without uploading it anywhere?

Yes. A local Whisper tool reads the video file from your own disk, extracts the audio and transcribes it on your machine, so the file never leaves it. The model itself is downloaded once, and after that no internet connection is needed. Online transcription services, by design, need the file uploaded to their servers first.

Should I export my transcript as TXT or SRT?

Export TXT if you want to read, quote or edit the words, and SRT or VTT if you need the timings, for subtitles or to find a moment in the video. They come from the same transcription, so you do not have to choose up front. Most Whisper tools export both formats from a single run.

Is Sablate free for transcription?

Yes, permanently and without an account, with limits. The free tier covers videos up to ten minutes, three jobs a month, the Fast and Balanced Whisper models, and TXT, SRT and VTT export. Your first longer video runs at any length. Pro is $29 one-time and removes the length and job limits and adds the Accurate and Best models.