Auto Subtitle Generator
Timed subtitles from speech, made by Whisper on your own device.
Free preview.
- Free preview: subtitles for the first half of the recording (up to 60 seconds), the first half of their words (up to 100), and the length of the whole recording.
- Locked until you unlock it: download and copy.
- Unlock: Pro pass, ₹179 for 30 days, a one-time payment that never renews.
Ways to unlock shows how to get the full result.
Printing this result is locked in the free preview.
Recording
Speech
The model is told this language; choose the one spoken in the recording.
Translates best with Standard or Most accurate.
Subtitle layout
Netflix’s limits are the start: 42 characters a line (16 for Chinese and Korean, 13 for Japanese, 35 for Thai), 2 lines, 5/6 of a second to 7 seconds. Changing them cuts the subtitles again at once, without transcribing again.
Choose how to run this
Stopping may still count toward today's free server AI.
Free server AI today (estimate):
Subtitles
Locked in the free preview. Opens the ways to unlock this result.
Locked in the free preview. Batch runs unlock with a pass.
Locked in the free preview. Query results unlock with a pass.
About the Auto Subtitle Generator
Choose a video or audio file, say which language is spoken, and get subtitles with times: the speech is recognised by Whisper, an open speech-recognition model, running in your browser on your own device. Your recording is not uploaded anywhere. Whisper writes the words with punctuation and times for each sentence; the generator then cuts them into subtitles that follow the Netflix guidelines — at most 42 characters a line in English, two lines, between 5/6 of a second and 7 seconds on screen, breaks after punctuation, and 2 frames between subtitles.
The result plays over your video at once, so you can check it as you watch. Change the line length, lines and durations and the subtitles are cut again without transcribing again. With a pass you save them as SRT, WebVTT, ASS, YouTube SBV, TTML or text; without one you get the free preview: the first part of the recording as timed subtitles.
How to use it
- Choose or drop a video or audio file (MP4, MOV, WebM, MKV, MP3, M4A, WAV, OGG, Opus or FLAC), up to 2 hours long.
- Choose the language spoken (it starts as your browser’s language). Tick Translate into English for English subtitles of speech in another language (with Standard or Most accurate: they translate far better than Fast).
- Pick a quality: Fast (Whisper tiny), Standard (Whisper base) or Most accurate (Whisper small). The first time, agree to the one-time download; its exact size is shown first.
- Press Generate subtitles and keep the tab open; a long recording takes a while. The subtitles appear over the video as they are made.
- Adjust Characters per line, Lines, Longest and Shortest if your platform needs other limits. With a pass, download the subtitles or copy them; without one, the free preview shows the first part and Ways to unlock shows how to get the rest.
Examples
jfk.mp4 — the end of John F. Kennedy’s inaugural address
1 00:00:00,170 --> 00:00:06,981 And so my fellow Americans ask not what your country can do for you, 2 00:00:07,064 --> 00:00:11,000 ask what you can do for your country.
One sentence becomes two subtitles that fit two lines of 42 characters and last under 7 seconds: the cut between them falls after the comma, and the line break keeps “ask not” together. Whisper tiny left out two commas (after “so” and “Americans”), so read the words through. Times inside a sentence are worked out from the length of its words.
Characters per line 25, Lines per subtitle 1
00:00:00,170 --> 00:00:01,728 And so my fellow 00:00:01,811 --> 00:00:03,534 Americans ask not 00:00:03,617 --> 00:00:05,996 what your country can do 00:00:06,079 --> 00:00:08,239 for you, ask what you 00:00:08,323 --> 00:00:11,000 can do for your country.
Five one-line subtitles for phones, Reels or Shorts, cut again from the same transcript without transcribing again.
Common uses
- Subtitles for YouTube, Vimeo, LinkedIn or a course platform from your own recordings.
- English subtitles for a talk or interview in another language (Whisper’s translation into English).
- Accessible captions for lectures and meeting recordings that must stay private.
- A first draft of subtitles to polish instead of typing them from scratch.
How the subtitles are made
- Sound: the recording is decoded on your device and mixed to 16 kHz mono; long silences are skipped, and speech is cut into parts of up to 30 seconds at pauses.
- Speech recognition: Whisper (open source; tiny 39 million parameters, base 74 million, small 244 million) was trained on 680,000 hours of speech in many languages (Radford et al., “Robust Speech Recognition via Large-Scale Weak Supervision”). It runs on your processor with WebAssembly, in a background worker, and writes punctuation and a start and end time for each sentence or phrase.
- Subtitles: sentences are cut into subtitles at the best places — sentence ends, then commas and other pauses, never right after “the” or “a” — so each fits the line and duration limits; lines are balanced and slightly bottom-heavy. Times inside a sentence are estimated from the length of each word, gaps under half a second are closed to 2 frames, and each subtitle stays up to half a second after the words end (Netflix timing guidelines).
- Line lengths start from Netflix’s guides: 42 characters for English (English guide) and the other languages written with spaces, 16 for Chinese and Korean, 13 for Japanese and 35 for Thai.
- Reading speed: the finished subtitles are checked against Netflix’s limit for adult programmes in their language, and the page says how many read too fast: 20 characters a second in English and Arabic, 17 in Spanish, French, German and most other languages (Spanish guide), 22 in Hindi, Marathi, Bangla, Tamil, Telugu, Kannada and Malayalam, 9 in Chinese, 4 in Japanese, 12 in Korean and 17 in Thai. A language Netflix has no guide for is checked against the English limit.
Choosing the quality
- Fast (Whisper tiny, about 44 MB): quickest; fine for clear speech in widely spoken languages.
- Standard (Whisper base, about 80 MB): better words and punctuation; the default.
- Most accurate (Whisper small, about 252 MB): noticeably better in noisy recordings and in languages with less training data, such as Hindi, Tamil or Bengali; slower, so download it on Wi-Fi.
Models are downloaded once from MySmartCoPilot, after you agree, and kept by your browser, so the next file starts at once. Fast is several times quicker than Most accurate, and phones are slower than computers; while it works, the page shows how much is done and, after the first part, the time left.
Limitations
- Choose the language spoken: the model is told the language and does not detect it here. With the wrong language the subtitles come out wrong or translated.
- Words can be misheard, especially names, numbers, technical terms, accents, music and people talking at once. Read the subtitles through before you publish them.
- Times are per sentence from the model and estimated within a sentence, so a subtitle may start or end a fraction of a second off; speakers are not told apart.
- Up to 2 hours and 2 GB per file; cut longer recordings into parts. AMR voice recordings must be converted first (for example with the Audio Converter).
- Translation is into English only (Whisper’s own translation) and is less accurate than transcription in the language spoken; the Fast model often translates badly, so choose Standard or Most accurate for it.
Privacy
Everything happens in your browser. What you enter or open here is not uploaded or stored by MySmartCoPilot. The first time, the speech model you choose (about 44, 80 or 252 MB) and its engine are downloaded from MySmartCoPilot after you agree, and your browser keeps them. Your recording never leaves your device.
Frequently asked questions
What do I get without a pass?
Without a pass, Auto Subtitle Generator shows subtitles for the first half of the recording (up to 60 seconds), the first half of their words (up to 100), and the length of the whole recording. Until you unlock it, the result can’t be downloaded or copied. A Pro or Premium pass, a one-time payment that never renews, unlocks the full result. The pricing page lists the passes and their prices.
Is my video uploaded?
No. The file is read and transcribed in your browser on your device. Only the speech model is downloaded, once, from MySmartCoPilot, after you agree.
Which languages can it subtitle?
The 99 languages Whisper was trained on, including English, Spanish, French, German, Portuguese, Hindi, Bengali, Tamil, Telugu, Marathi, Urdu, Arabic, Chinese, Japanese and Korean. Accuracy depends on how much training speech a language had and on the quality you choose: English and widely spoken European languages give usable results with every quality; for Hindi and other Indian languages choose Most accurate.
Why do I have to choose the language?
Telling the model the language gives far better results, and this on-device engine needs it: it does not detect the language by itself. The list starts with your browser’s language.
What does the free preview include?
Without a pass the generator transcribes the first half of your recording, up to 60 seconds (ending at a pause, so no word is cut), and shows the first half of its words, up to 100, as timed subtitles over your video, together with the length of the whole recording. Downloading, copying and opening the subtitles in the editor need the full result: a Pro pass, or an unlock of this one result, transcribes the rest on your device from the file still on the page.
Can I change the subtitles after they are made?
Change the line and duration limits on this page and they are cut again at once. To change words or times, download the subtitles (SRT or WebVTT) and edit them in the subtitle program you use.