Local ConverterAll tools

Transcribe audio or video locally

Turn spoken audio into text or timed subtitles with offline Whisper. Use Transcribe audio or video locally locally with practical instructions, output settings and format.

A practical guide

Transcribe audio or video locally

Turn a short audio or video excerpt into text with the bundled local Whisper tiny model. Choose plain text, SRT or VTT, and use automatic language detection or a supported language code. The selected audio is limited to 120 seconds, and every transcript needs review for names, timing and recognition errors.

  • Draft notes from a short voice memo
  • Create initial subtitles for a video excerpt
  • Transcribe a brief spoken explanation

How to use Transcribe audio or video locally

  1. Choose a recording with speech

    Upload one supported audio or video containing an audio track. Clear speech with limited background noise gives the small local recognition model a better starting point.

  2. Select the speech excerpt

    Enter start and length in seconds, with a maximum selected length of 120 seconds. Choose the relevant passage rather than expecting transcription of a complete long recording.

  3. Choose language and text format

    Select TXT, SRT or VTT and use auto language detection or a supported language code. Run the tool with the bundled Whisper tiny model available locally.

  4. Correct the transcript

    Download the text or subtitles and review them while listening. Correct names, numbers and missed words, and check cue timing before using the output as captions.

Understand the settings

Transcript format
Choose transcript format according to the result you need. Available choices: TXT; SRT; VTT. The initial selection is TXT.
Language code (auto detects)
Language code (auto detects) is applied to this tool job. Use the value or syntax described in the example below. The initial value is “auto”.
Start (seconds)
Start (seconds) accepts values from 0 to 7200, in increments of 0.1. The initial value is 0. Change it for a specific output requirement, then check the result before processing more files.
Length (seconds)
Length (seconds) accepts values from 0.1 to 120, in increments of 0.1. The initial value is 20. Change it for a specific output requirement, then check the result before processing more files.

Draft captions for a short explanation

A sixty-second video explanation needs initial captions. Select the relevant minute, choose SRT and the known language code, then transcribe. Listen while editing the downloaded cues, correcting technical names and checking each timing boundary before sharing the captioned version.

Check your result

Upload a file, adjust settings and run the tool. All processing uses the local engine. Requires the bundled local Whisper tiny model. Maximum 120 seconds. Recognition can be inaccurate; review the transcript. Open the saved output in the app or workflow where you plan to use it. Check one representative result before repeating the same settings with more files. The original files stay unchanged.

What to know before you start

  • The first use downloads the local Whisper tiny model. Speech stays on your device; generated words and timings need review.
  • Files are processed on this device. Browser memory and engine-specific limits apply. Save results before reloading.
  • Requires the bundled local Whisper tiny model. Maximum 120 seconds. Recognition can be inaccurate; review the transcript.
  • Upload one file with a combined size of at most 100 MB. The completed output must fit within 200 MB.
  • Files are checked by the local engine; a matching filename alone is not proof of valid contents.
  • Selections and completed jobs stay available while you navigate inside this tab. Use My files to revisit results. Reloading or closing the tab clears the workspace; download results you want to keep first.

Transcribe audio or video locally: common questions

How much audio can be transcribed?

The selected segment is limited to 120 seconds. A longer supported source can be used to select a short passage, but this workflow does not automatically divide and transcribe an entire long recording in one operation.

Are subtitle times relative to the source?

Recognition runs on the extracted audio segment, so SRT and VTT timestamps are relative to that excerpt’s beginning. If you need captions aligned to the original source timeline, adjust the resulting times by the excerpt’s start offset.

Why can names or numbers be wrong?

Whisper tiny is a small recognition model, and accents, background noise, overlapping speech and unfamiliar terms can produce errors. Treat the transcript as a draft and verify meaningful names, quantities and statements against the actual recording.

Does the tool work without the local model?

No. It requires the bundled Whisper tiny model files, and reports an error if they are unavailable. Language auto-detection and transcription run through that local model rather than falling back to an unconfigured remote service.

Does Transcribe audio or video locally upload files to another service?

Processing runs in this browser, on your device. Files and pasted input are not uploaded to a conversion server. The website and required engine assets must load first; large operations depend on available device memory.

Can I continue after changing pages?

Selections and completed jobs stay available while you navigate inside this tab. Use My files to revisit results. Reloading or closing the tab clears the workspace; download results you want to keep first.