About VoiceToTextFree
VoiceToTextFree is an independent, browser-based subtitle and transcript generator maintained by Abdul Samad. The goal is simple: make a useful subtitle tool that does not require an account, an upload or a per-minute payment, and explain its limits honestly.
Why this site exists
Most online subtitle generators send your recording to a server. That is a reasonable business model, but it is a poor fit for interviews, legal recordings, medical notes, student conversations and anything covered by a confidentiality agreement. Modern browsers can run a capable speech model locally. VoiceToTextFree is built around that capability, and every design decision follows from it: no backend, no file storage and no account.
The site also exists because subtitle advice online is often vague. A page that says “use a good subtitle format” without showing the timestamp syntax, the character limits or the platform's current rules is not much help. The guides use worked examples, real timestamps and links to primary sources. Where a fact depends on a platform that changes its menus, we say when it was checked and link the help page.
How it works
The tool uses ONNX builds of OpenAI Whisper through transformers.js. Audio is extracted with ffmpeg.wasm at 16 kHz mono, split into overlapping chunks, and transcribed by a model running with WebGPU where the browser supports it or WebAssembly otherwise. The transcript is editable, and exports are generated in the same tab. The first run downloads the model; the browser caches it for later runs. The full technical description and the sources for model sizes and language coverage live in the Whisper guide and the corrections log.
Who maintains the site
Abdul Samad is the creator and maintainer of VoiceToTextFree and is responsible for the site's browser-based tool, subtitle guides and corrections. All original tooling, editorial content and maintenance work on this site are credited to him unless a page states otherwise. The site does not publish invented credentials, client lists or testimonials. If you need to know who is responsible for a claim, the author box on every article and this page are the record.
Contact: ihsanaslamm@gmail.com.
Honest limitations
- Automatic transcription is a draft, not a certified transcript. Names, numbers and punctuation need review.
- Accuracy varies by language, accent, microphone and background noise. Low-resource languages can be weak.
- Files are limited to 500 MB and processing happens in memory, so long recordings can exceed a phone's capacity.
- The tool translates only into English (Whisper's translate task). It is not a general subtitle translator.
- Offline use requires the model to have been downloaded once in the same browser profile.
- Video burn-in re-encodes the file and is slower than generating a sidecar subtitle file.
How to reach us
Email ihsanaslamm@gmail.com for corrections, privacy requests or questions. The contact page explains what to include and the expected response time. If you find a factual error, the corrections log shows how previous fixes were handled.