# Subtitle Generator

> Free AI Audio/Video → SRT & VTT • Whisper Large-v3 • 18 Languages

Free AI subtitle generator. Convert audio or video to SRT or VTT with Whisper large-v3. 18 languages, no signup, no watermark, no daily limit.

URL: https://tools.scoreroute.com/tools/subtitle-generator/

Markdown: https://tools.scoreroute.com/tools/subtitle-generator/.md

## 1. Upload Audio or Video


## 2. Transcription Options


## 3. Generate


### Cue Preview


### Subtitle Text


## How It Works

- Drop your audio/video file (up to 25 MB).
- Pick a Whisper model and language (or auto-detect).
- Click **Generate Subtitles** - Whisper transcribes with millisecond-precise timestamps.
- Preview cues, download SRT/VTT, or open in our [Subtitle Editor](/tools/subtitle-editor/).


## Common Use Cases

- YouTube creators adding captions to videos
- Students transcribing recorded lectures
- Podcasters generating episode transcripts
- Course creators producing accessible lessons
- Translating audio interviews for journalism
- Indexing video libraries for search


## Limits & Caveats

- Groq Whisper accepts files up to **25 MB**
- Long audio (over ~30 min) may hit Groq's 60 RPM rate limit
- Processing time is typically 0.3-0.5x audio length
- Background noise / heavy accents may reduce accuracy


### About the Subtitle Generator

Hardcoded subtitles are the most reliable way to make video accessible, but creating them by hand is slow and error-prone. This tool turns the audio track of any video - or any standalone audio file - into properly time-stamped SRT or VTT captions in one click. The same Whisper model powers commercial services that charge per minute, but here it runs through Groq's free developer tier so you can use it as much as you need.

The output uses standard SubRip (SRT) or WebVTT (VTT) format. SRT is the format YouTube, VLC, Handbrake, and most NLE editors accept on import. VTT is the format used by HTML5 video tags, JW Player, and the BBC iPlayer-style web players. Both formats use the same time-aligned cue structure (HH:MM:SS,mmm index text), so you can transcribe once and publish anywhere.


### Why Use This Instead of a Chatbot

Large language models can reformat text into a cue block, but they cannot listen to audio. They also hallucinate timestamps that don't match the actual speech, so the caption flashes on screen at the wrong moment or lingers after the speaker has moved on. A dedicated speech recognition model produces frame-aligned timing, which is what viewers actually need. The output is structured, deterministic, and ready to ship.


### What to Do After Generating

Open the file in our [Subtitle Editor](/tools/subtitle-editor/) to fix any name spellings, brand-specific terms, or punctuation Whisper got wrong. The editor validates cue timing (no overlaps, no negative durations), supports batch shift, split-long-cue, and merge-adjacent operations, and exports back to SRT or VTT with millisecond precision.


### Related Tools


#### Subtitle Editor

Fix timing, split cues, validate overlaps, export SRT or VTT. 100% browser-based.


#### Audio Cutter

Trim audio files before transcribing to stay under the 25 MB limit.


#### Audio Compressor

Reduce file size to fit longer audio under the 25 MB cap.


#### Audio to Text

Real-time browser speech recognition for fully offline transcription.


#### Video Cutter

Trim video files before uploading for subtitle generation.


### Frequently Asked Questions

**What audio and video formats are supported?**

The tool accepts MP3, WAV, M4A, OGG, FLAC for audio and MP4, MOV, WebM, MKV for video. Files up to 25 MB are accepted; video files are processed by extracting the audio track on the Groq side using Whisper, which scans the audio stream directly so no client-side decode is needed.

**How accurate is the AI transcription?**

The tool uses OpenAI Whisper large-v3 - the highest-accuracy public speech model. On clean English audio you can expect 95-98% word accuracy; on noisy or accented speech, 85-95%. The faster turbo model is roughly 5x faster with a small accuracy trade-off, and the distil model is English-only but the fastest of the three.

**Which languages can the tool transcribe?**

The tool supports 18 languages: English, Chinese, Spanish, French, German, Japanese, Korean, Portuguese, Italian, Russian, Arabic, Hindi, Dutch, Polish, Turkish, Vietnamese, Indonesian, and Thai. Leave the dropdown on Auto-detect and Whisper will identify the spoken language automatically - useful for mixed-language content.

**What is the difference between SRT and VTT?**

SRT (SubRip) is the most widely used subtitle format and works in VLC, YouTube uploads, Handbrake, and most video editors. VTT (WebVTT) is the standard for HTML5 video players and BBC-style web players. The tool lets you export either format from the same transcription - just switch the Format dropdown before downloading.

**Is there a file size limit?**

Yes - 25 MB. The Groq Whisper endpoint rejects larger files. If your audio is longer than about 30 minutes at typical bitrate, trim or compress it first using our Audio Cutter or Audio Compressor. For very long content, split it into 25 MB chunks and transcribe each one.

**Can I edit the generated subtitles afterwards?**

Yes. The tool includes an "Open in Subtitle Editor" button that loads the generated SRT or VTT directly into our companion Subtitle Editor, where you can fix typos, adjust timestamps, split or merge cues, validate overlaps, and export back to either format. Everything stays in your browser.

**Is my audio uploaded to a server?**

Unlike most of our tools, this one DOES upload your audio/video to our Cloudflare Function, which forwards it to Groq Whisper for transcription. The file is processed transiently and not stored on either end - both Groq and our function are stateless. If you need a fully offline alternative, see our browser-based Audio to Text tool which uses Web Speech API.

**Why can't AI chatbots do this task?**

ChatGPT, Claude, and Gemini can reformat text into timestamped subtitle blocks, but they cannot listen to audio or video. They also cannot guarantee millisecond-accurate timestamps aligned to actual speech - a hallucinated timestamp breaks the entire viewer experience. This tool uses a dedicated speech recognition model so the timing matches the audio exactly.

## References

- [OpenAI — Whisper (Speech-to-Text)](https://openai.com/research/whisper)
- [Matroska — SubRip (SRT) Subtitle Format](https://www.matroska.org/technical/subtitles.html#srt-subtitles)
- [W3C — WebVTT (Web Video Text Tracks)](https://www.w3.org/TR/webvtt/)