Transcribe audio and video files to text with OpenAI Whisper or the GPT-4o transcribe models and your own API key. Upload a file or record straight from your microphone.
Use your own API key to get started.
An OpenAI API key is required
This tool transcribes with OpenAI: Whisper and the GPT-4o transcribe models. Other providers (Gemini, Anthropic, Z.AI, Kimi, and so on) cannot transcribe audio. Click "Use Your API Key" above and add an OpenAI key, or see the BYOK tools page for how bring-your-own-key works. The key is stored in your browser only.
Privacy difference worth knowing: most tools on this site process files locally in your browser. This one does not. Your audio IS sent to OpenAI for transcription, using YOUR API key, and is processed under OpenAI's data policy. Nothing is stored by Kit Utility. Read more on the BYOK tools page.
MP3, WAV, M4A, OGG, WebM audio, or a video file (only its audio track is used). Max 25 MB.
Recording uses your microphone (the browser will ask for permission) and keeps the audio in memory until you transcribe it.
Picking the language improves accuracy slightly. Leave on auto-detect for mixed-language audio.
Modern speech recognition is unusually good at messy real-world audio: accented speech, background noise, crosstalk, and quiet recordings that defeat simpler engines. This tool puts OpenAI's three transcription models to work: Whisper, GPT-4o Transcribe, and the cheaper GPT-4o mini Transcribe, all selectable in the model picker. Upload an audio file (MP3, WAV, M4A, OGG, WebM) or a video file, and its audio track is transcribed directly (handy for turning interviews, meeting recordings, lectures, podcasts, and voice memos into searchable text). Or skip the file entirely and record from your microphone, then transcribe the recording in one click.
This is a bring-your-own-key (BYOK) tool: it needs an OpenAI key specifically, and you pick the model yourself. Gemini, Anthropic, Z.AI, and the other providers on this site generate text and images but cannot transcribe audio, so they do not appear in the provider list. Your key is stored in your browser only, and requests pass through our proxy to OpenAI under your account. That leads to the honest privacy note: unlike this site's FFmpeg video tools, where files never leave your device, here your audio IS uploaded to OpenAI for processing. Kit Utility stores nothing; OpenAI's API data policy governs retention.
Transcripts come back as plain text with punctuation, ready to copy or download as a .txt file. For long recordings remember the 25 MB upload cap (about 30 minutes of MP3, less for WAV). Need the opposite direction? The Text to Speech tool reads text aloud locally and free. To pull just the audio out of a video first, use Audio Extractor, and to remove audio instead, see Video Muter.
Transcription runs on OpenAI models: Whisper, GPT-4o Transcribe, and GPT-4o mini Transcribe, all reachable only with an OpenAI API key. The other providers supported on this site (Gemini, Anthropic, Z.AI, Kimi, MiniMax, DeepSeek, Clipdrop) have no audio transcription capability, so a key from any of them cannot power this tool.
Yes, and that is the key privacy difference from the local FFmpeg tools on this site. Your audio is sent from your browser through our proxy to OpenAI using your own API key, transcribed there under the OpenAI API data policy, and the text is returned. Kit Utility does not store your audio, transcript, or key. If your recording is sensitive enough that uploading it anywhere is a problem, do not use this tool.
Any common audio format (MP3, WAV, M4A, OGG, WebM) plus video files such as MP4, MOV, and MKV, from which the audio track is used. The hard limit is 25 MB, checked in your browser before anything uploads. Roughly: 25 MB is about 30 minutes of MP3 at 128 kbps, or 2 to 3 minutes of uncompressed WAV.
Whisper is the default and a strong all-rounder: it handles accents, background noise, and technical vocabulary well, and adds punctuation automatically. GPT-4o Transcribe is newer and often more accurate on difficult audio, and GPT-4o mini Transcribe is the cheaper, faster variant of it. Expect near-perfect text for clean single-speaker audio, and good-but-imperfect text for heavy crosstalk or very noisy recordings. Choosing the spoken language improves results slightly over auto-detect.
No. Recording happens locally with the browser MediaRecorder and the audio stays in memory (with a local preview you can replay). It is only uploaded when you click Transcribe. Closing or reloading the page discards the recording without any network transfer.