Video to SRT & MP4 to SRT Converter — Generate AI Subtitles Online

Key takeaways

• •AI generates .srt subtitle files from raw video — no manual audio extraction step.
• •Accepts MP4, MOV, MKV, WebM, AVI directly; max file size 5 GB on VexaScribe.
• •Output is timestamped, line-broken, UTF-8 encoded — drops into YouTube, Premiere, DaVinci, VLC, CapCut without conversion.
• •Accuracy 92-97% on clean audio; review proper nouns and technical terms before publishing.
• •Cost $0.20-$0.60 per video hour AI; $90-$300 per video hour human (Rev, 3PlayMedia).
• •Processing 5-15 minutes per video hour AI; 12-48 hours human turnaround.
• •Diarization is optional — useful for interviews and panels, irrelevant for single-speaker explainers.

How to convert video to SRT (4 steps)

1. 1

   Upload the video file

   MP4, MOV (iPhone default), MKV, WebM, AVI up to 5 GB per file (approximately 8-10 hours of compressed video). Audio is extracted from the video container automatically — no manual conversion to MP3 or WAV required. Free trial accepts the first 30 minutes of any file.

2. 2

   Choose source language and diarization

   Select source language from 99 supported languages, or use auto-detect for clean monolingual audio. Toggle speaker diarization on for multi-speaker video (interviews, panels, podcasts) — labels speakers as Speaker 1, Speaker 2 for renaming later. Diarization is included on every paid plan with no tier gating.

3. 3

   Wait for processing

   AI runs at 4-10× real-time. A 30-60 minute video processes in 5-15 minutes. VexaScribe emails you when the SRT is ready. While waiting, you can queue additional video uploads — useful for batch caption generation across a YouTube channel or course.

4. 4

   Download the .srt and review

   Download the .srt file (UTF-8 encoded, timestamped, line-broken to ~42 chars/line). Quick proofread for proper nouns, brand names, technical terms (5-15 min/hr). Drop into YouTube Studio (Subtitles → Upload file), Premiere Pro (Captions panel → Import), DaVinci Resolve, CapCut, or VLC. DOCX and TXT also exported for downstream workflows.

Supported video formats

VexaScribe accepts the most common video containers directly — no manual audio extraction step. The platform extracts the audio track using ffmpeg internally and routes the audio through the Whisper Large-v3 transcription pipeline.

Max file size: 5 GB per upload — covers approximately 8-10 hours of typical compressed video at 720p-1080p. For longer files, split with a free tool like LosslessCut before uploading.

Self-hosted Whisper users must extract audio first since base Whisper accepts only audio formats: ffmpeg -i input.mp4 -vn -acodec libmp3lame audio.mp3, then whisper audio.mp3 --output_format srt.

What is an SRT file?

SRT (SubRip Subtitle) is the oldest and most widely supported subtitle format. It originated with SubRip, a freeware Windows tool from the early 2000s that "ripped" subtitles from DVDs — the format was designed for simplicity: a plain UTF-8 text file with numbered blocks, each containing an index, a timestamp range, and one or two lines of caption text. That simplicity is why it outlasted every competitor: no binary encoding, no proprietary spec, no renderer required. It's natively supported by YouTube, Vimeo, Premiere Pro, DaVinci Resolve, CapCut, VLC, Windows Media Player, every social platform, and every streaming service.

1
00:00:00,000 --> 00:00:03,500
Welcome to the introduction lecture.

2
00:00:03,500 --> 00:00:07,200
Today we'll cover three main topics.

3
00:00:07,200 --> 00:00:11,800
Let's start with the first one.

Anatomy of a block:

• →Index: sequential integer starting at 1. Must be unique and ordered.
• →Timestamp range: HH:MM:SS,mmm --> HH:MM:SS,mmm. Note the comma (not period) before milliseconds — common cross-platform compatibility issue.
• →Text lines: one or two lines per block. Broadcast standard caps at 42 characters per line; mobile-friendly drops to 32-36.
• →Blank line: separates each block. Critical — missing blank lines break parsing in some players.
• →Encoding: UTF-8 without BOM is the safe default. UTF-8 with BOM works in most players but trips a few legacy ones.

4 formatting mistakes that break SRT playback:

• 1.Period instead of comma in timestamps. 00:00:03.500 instead of 00:00:03,500. VLC and most desktop players tolerate this; YouTube Studio does not — upload fails silently with no error message.
• 2.Missing blank line between blocks. Some editors strip trailing blank lines on save. Result: the parser reads two blocks as one, producing doubled subtitles with wrong timing.
• 3.UTF-8 BOM on byte 1. The BOM (EF BB BF) is treated as a text character on line 1 by some parsers, so the first subtitle index becomes 1 instead of 1. Fix: save as UTF-8 without BOM (Notepad++ → Encoding → UTF-8 without BOM).
• 4.Overlapping timestamps. A block that starts before the previous one ends. Some players merge the cues; others drop one. Generated by AI when VAD boundaries are too aggressive — VexaScribe enforces a 10ms minimum gap between cues.

SRT vs other subtitle formats:

Verdict: Use SRT for 99% of workflows — video editors, YouTube, social media, streaming. VTT only when you need positioned captions on a custom HTML5 player. ASS/SSA for anime subtitles. TTML for broadcast delivery specs.

How VexaScribe builds readable cues

Transcription accuracy is one half of a usable SRT. The other half is cue quality — splitting long speaker turns into subtitles that read naturally on screen and stay in sync with speech. A 30-second monologue can't be one subtitle. A two-word reaction shouldn't flash on screen for a quarter of a second. VexaScribe handles this with a word-level cue splitter built specifically for editor-ready output.

• ● Word-accurate timing. Cue start and end times come from the engine's word-level timestamps, not linear interpolation across a segment. Subtitles appear when speech starts and disappear when it ends.
• ● 80-character / 5-second target. The readable-web-subtitle range used by Descript, Sonix, and Vimeo. Tighter than broadcast TV (42 chars), looser than YouTube auto-captions. Hard ceiling of 10 seconds for compatibility with every standard player.
• ● Natural splitting points. The splitter prefers sentence-ending punctuation (. ! ?), then commas, then word boundaries — never mid-word.
• ● Dramatic pauses preserved. When a speaker pauses 2-3 seconds mid-thought (a beat in a speech, a breath in an audiobook), the cue holds across the silence rather than producing a sub-second flash followed by a blank screen.
• ● Speaker labels carry through. If diarization is on, every cue keeps the [SPEAKER_00] prefix, honoring per-word speaker changes within a segment when present.

The result: an SRT that drags into Premiere Pro, Final Cut Pro, DaVinci Resolve, or CapCut and works without manual cleanup. For deeper detail on the splitting algorithm and Netflix/BBC timing standards, see our SRT generator page.

Accuracy by video type

Whisper Large-v3 (the model VexaScribe uses) hits 95-97% accuracy on clean single-speaker video but degrades predictably with audio conditions. Plan your review time based on the type of video you're subtitling.

Where AI consistently misses: proper nouns (brand names, product names, technical terms) at 20-30% error rate even on clean audio; numbers spelled vs digits ("twenty twenty six" vs "2026"); homophones (their/there/they're); rapid-fire counts and lists. Always proofread before publishing public-facing captions.

For accuracy methodology, see how accurate is Whisper? with WER benchmarks across LibriSpeech and FLEURS.

Cost: per-video and bulk math

AI subtitle generation is genuinely cheap — typically $0.20-$0.60 per video hour on consumer apps. Human captioning is 150-1,500× more expensive. The cost math only flips toward human if you specifically need verbatim, court-grade, or ADA-certified broadcast captions.

Bulk math example: a YouTube channel publishing 4 videos/month at ~15 minutes each = 1 hour of new video monthly. SRT generation costs $0.20-$0.60 on AI versus $90-$300 with human captioning — a 150-1,500× difference. For a 40-episode course (~20 hours of video total), AI runs $4-$12 versus $1,800-$6,000 human.

For full cost analysis across the 14-tool transcription market, see how much does transcription cost? with verified 2026 pricing and an interactive calculator.

Multi-language SRT (99 languages)

VexaScribe supports SRT generation in 99 languages via Whisper Large-v3, including all major European, East Asian, and Middle Eastern languages — Spanish, French, German, Italian, Portuguese, Japanese, Korean, Mandarin, Arabic, Russian, Hindi, Turkish, Vietnamese, Polish, Dutch, Swedish, plus 83 more. Source language is auto-detected from the audio or manually selectable.

Common workflows: Spanish-language creator generates Spanish SRT for native audience plus English SRT for international viewers. Japanese tutorial creator generates Japanese + English + Spanish SRTs for global reach. Documentary filmmaker generates SRTs in source language plus festival-circuit target languages (English, French, German). See also transcribe and translate audio.

Common SRT errors and fixes

Most SRT problems come from one of five issues. Here's how to recognize and fix each one.

Recommended free editor: Subtitle Edit (Windows) — handles framerate conversion, line balancing, encoding fixes, sync drift, and translation memory in one tool. For cross-platform editing, Aegisub. For quick edits, the built-in caption editors in YouTube Studio, Premiere Pro, and DaVinci Resolve all work.

How to create an SRT file for a video

The 4-step workflow above is the tool version. Here's the version that answers "how to create an SRT file for a video" as a general question — including the DIY-with-Whisper path if you don't want to use a cloud tool.

Path A — Cloud AI (fastest, easiest)

1. Upload video to a cloud AI transcription tool (VexaScribe, Descript, TurboScribe, Sonix) OR paste a YouTube/direct URL.
2. Select source language (auto-detect works for clean monolingual audio) and toggle speaker diarization on for multi-speaker video.
3. Wait 5-15 minutes per video hour for AI processing.
4. Review the auto-generated .srt in the built-in editor — fix proper nouns (10-20% error rate on brand names), split lines over 42 characters, adjust any timing that drifts.
5. Download the .srt (UTF-8 encoded). Upload to YouTube or import into Premiere/DaVinci/CapCut.

Path B — Self-hosted Whisper (free, technical)

1. Install Python 3.10+ and OpenAI Whisper: pip install openai-whisper
2. Extract audio from video with ffmpeg: ffmpeg -i input.mp4 -vn -acodec libmp3lame audio.mp3
3. Transcribe with SRT output: whisper audio.mp3 --model large-v3 --output_format srt
4. Whisper writes audio.srt to the current directory. Rename to match your video filename for auto-loading in VLC and most players.
5. Note: base Whisper doesn't include speaker diarization. Pair with pyannote.audio 3.1 or use WhisperX for combined ASR + diarization + word-level alignment.

Path C — YouTube auto-captions (free, English-primary)

Upload the video to YouTube as unlisted, wait 2-6 hours for YouTube auto-captions to generate, then in YouTube Studio go to Subtitles → download the .srt via the three-dot menu. Accuracy runs 82-92% for English, lower for other languages. Adequate for accessibility drafts; not adequate for broadcast-quality captions without heavy editing.

Video to SRT vs alternatives

We position VexaScribe honestly: it's the right pick for most batch SRT workflows (direct video upload, 99 languages, speaker labels, $2/mo entry). Other tools win specific lanes — here's the honest read.

When to pick something other than VexaScribe. If you're editing the video and need captions integrated with the editor, Descript is the right call — the caption editor sits inside the video editor. If your channel is English-only and you don't care about accuracy quality, YouTube auto-captions are free and require zero workflow steps (the captions appear automatically after upload). If you have technical skills, a GPU, and high-volume needs, self-hosted Whisper is free forever — pay the setup cost once, run unlimited.

See also SRT generator (format deep dive) and transcription tool alternatives.

Embedding the SRT in your video

Once you have the .srt file, embedding it is straightforward. Quick reference for the most common destinations:

For step-by-step embedding tutorials across platforms (including iPhone, Final Cut Pro, and burning subtitles permanently into video), see how to add subtitles to a video.

FAQ

Methodology & disclosure