Transcribe an Uploaded File
Upload an audio or video file and get a full transcript with speakers, languages, and optional translation.
Already have a recording? Upload it to Kalima and get a full, speaker-labeled transcript you can read, edit, search, and export, with no need to play it back live. This page walks you through the upload flow from picking a file to opening your finished transcript.
What you can upload
Kalima accepts most common audio and video files. Audio formats include MP3, WAV, M4A, AAC, FLAC, OGG, Opus, WebM, AIFF, and AMR, and WhatsApp voice notes work too. Video formats include MP4, MOV, and WebM. For video, only the audio track is transcribed. Kalima extracts the audio in your browser, so the video itself is never uploaded or stored.
Speaker separation and language detection run on every upload, so there is nothing to switch on. Different voices are separated for you, multi-language recordings are supported, and the spoken language is detected automatically. For the full list of formats and the size, duration, and weekly limits that apply to your plan, see Supported Formats & Limits.
How to transcribe a file
Open the upload card. Go to the Studio and choose the "Upload audio or video file" option from the new-session view.
Choose your file. Drag a file onto the card, or click Browse to pick one from your device. You can play it back first. The built-in preview player has a waveform, volume control, and playback speeds from 0.5x to 2x, so you can confirm you grabbed the right recording before spending any balance.
Set your options (all optional). Add expected languages, turn on translation, and add context to improve accuracy. These are covered in detail below.
Check the cost. Before you start, Kalima estimates the time cost of the file and shows it alongside your remaining balance, so there are no surprises.
Start the transcription. Click "Start Transcription". The upload card switches to a staged progress view: extract audio (for video), upload, prepare, and transcribe.
Open your transcript. Once the upload finishes, you can leave the page, and processing continues in the background. Return any time to find the completed transcript in your session history.
You can leave the page only after the upload itself has finished. If you navigate away while the file is still transferring, the upload is cancelled and you will need to start it again.
Improve accuracy before you start
Three optional settings in the upload panel help Kalima get your recording right. You can use any combination of them.
Expected languages
Tell Kalima which language or languages to expect in the recording. This sharpens results for accented speech, less-common languages, and conversations that switch between languages. Leave the field empty and Kalima detects the language automatically, so this is a helpful nudge, not a requirement. Add only the languages actually present for the best results. Learn more in Languages & Detection.
Translation (2-for-1)
Turn on the "Translation (Optional)" switch and pick a language under "Translate to" to get a translated transcript in the same pass as transcription. The result holds both the original and the translation, aligned by timing.
Translation on a file upload is 2-for-1. It is processed alongside transcription at no additional time cost. You are billed only for the audio duration, whether or not translation is on.
You can translate into one target language per upload. The translation default follows your preferred primary language, and you can change it before you start. See Live Translation for more.
Context (names, terms, jargon)
Open "Context (Optional)" to give Kalima background about the recording: a short description, plus key terms and proper nouns such as people's names, product names, and acronyms. When you are translating, you can also specify how you want certain terms translated. This helps Kalima spell specialized words correctly instead of guessing them phonetically. An "Active" badge appears once context is set.
Context is ideal for medical, legal, technical, or product recordings, and you can save up to 2,000 characters. See Improve Accuracy with Context for tips on writing effective context.
Cost and billing
You are charged from your time balance based on the actual transcribed duration, never a flat estimate.
Before you start, Kalima estimates the file's duration and compares it to your balance, showing the cost and what you will have left. If your balance is too low, you will see an "Insufficient Balance" warning with a button to add balance or upgrade.
When you start, an estimated amount of time is reserved automatically.
When it finishes, your balance is reconciled to the exact duration that was transcribed. Any over-reservation is refunded, or a small difference is charged if the file ran longer than estimated.
If you cancel an upload, or it fails, the reserved time is released back to you. For how minutes and credits work, see Minutes, Credits & Balance, and for the full picture of background jobs and error handling, see Processing & Costs.
What happens after you start
Once your file is uploaded, transcription runs on Kalima's side. The session appears in your history as "Processing" and updates to "Completed" when it is done, so you can close the tab, work on something else, and come back to a finished transcript. Longer recordings naturally take longer to finish; multi-hour files can take several minutes or more.
If something goes wrong, Kalima tells the difference between temporary problems (such as a busy service or a timeout) and permanent ones (such as an unsupported format or a file that is too large). Temporary issues are safe to retry; permanent ones need a change to the file or your plan. See File Upload Issues for the full problem-to-fix list.
Common uses
Frequently asked questions
Audio: MP3, WAV, M4A, AAC, FLAC, OGG, Opus, WebM, AIFF, and AMR. Video: MP4, MOV, and WebM, where the audio is extracted automatically. WhatsApp voice messages (.opus) are supported. If a file is not supported, the card shows "Unsupported file format" and lists the accepted types.
A file can be up to 5 hours (300 minutes) of audio, enforced on every plan. Your maximum file size and per-file duration also depend on your plan. See Supported Formats & Limits for the full table.
No. Once the file finishes uploading, processing continues in the background and you can leave the page. Come back any time to find the finished transcript in your sessions. If you leave while the file is still uploading, the upload is cancelled.
You are charged from your time balance based on the actual transcribed duration. Kalima shows an estimated cost and your remaining balance before you start, reserves time when you begin, and reconciles to the exact amount when it finishes, refunding any difference.
No. When you enable translation on a file upload, it is processed alongside transcription at no additional cost. You can translate into one target language per upload.
Yes. Speakers are separated automatically on every upload, and you can rename them afterward in the transcript editor.
No. The language is detected automatically, and mixed-language recordings are supported. You can optionally add expected languages to improve accuracy.
The video may have no audio track, or the audio is silent or too quiet. Confirm the video has audible speech and try a version with the audio intact.