Confirm authority and purpose
Before uploading, confirm that the client or recording owner permits transcription by an external service. Decide who may access the audio and how the source text will be retained.
A preparation tool for language professionals
Create a source-language transcript before you interpret
Turn a saved recording into searchable text in the language that was spoken. Review names, terminology, numbers and speaker turns while keeping the audio available as the authoritative source.
From recording to reviewed source text
Transcribe what was spoken
Upload the saved audio or record it in the app. Choose a language when you know it, or use automatic detection. The result stays in the source language rather than being converted into a target language.
Resolve the risky details
Search for names, organisations, dates, numbers and repeated technical terms. Compare each consequential passage with the audio and correct the transcript before using it for preparation.
Export or share deliberately
Copy text, download a TXT file, or create an expiring transcript link that can later be revoked. Share only the minimum material allowed by the assignment and your confidentiality obligations.
A preparation checklist for names, numbers and uncertain terms
Before processing
Confirm the permitted purpose, external-provider use, audience and retention rules for this assignment. Gather approved names and reference material. If the source is too sensitive for this workflow, use the client's permitted process instead.
Resolve from evidence
Mark the timestamp and the uncertain phrase. Replay the audio, then compare an approved client reference. If neither resolves a name or number, keep it unresolved and ask the client. A plausible transcript is not a terminology decision.
Synthetic example
An invented draft says “fifteen units.” The passage is unclear on replay and a reference lists a different quantity. Record the conflict and ask; do not silently replace the spoken number with the reference value. This example is not a practitioner case or a quality benchmark.
Keep the review trail
The blank sheet has fields for time, source term, uncertainty, approved reference, audio check, client question, review state and sharing or retention decision. Fill it in manually; the product does not create or verify a glossary.
Download preparation worksheet (CSV)
No account is required. Completed worksheets stay on your device. This is a possible preparation method, still awaiting practitioner feedback, not a universal professional requirement.
Useful where preparation meets imperfect audio
Speaker turns you can repair
Automatic speaker separation is optional and may mislabel overlap or similar voices. Rename a detected speaker and reassign an individual turn, then verify attribution against the original recording.
Right-to-left source text
Hebrew, Arabic, Farsi and Yiddish transcripts are laid out in their natural reading direction, including mixed-language passages. This improves navigation; it is not a promise that every proper noun or borrowed term is correct.
Vocabulary before the recording
Add client names and technical terms under “Words and names to recognize” before transcription. That context is used only to improve the transcript; it does not create a glossary or replace checking terminology against the recording and the client's approved references.
What this saved-recording workflow does not do
This workflow does not translate or interpret the recording, join a live assignment, assess meaning in context, or produce a certified transcript. It does not know the client's glossary or the consequences of a disputed word. The language professional remains responsible for the preparation method and final output. This page describes a possible workflow, not a universal professional rule; assignment requirements, ethics codes, consent duties and local law can differ.
The separate Windows Hebrew-English live-caption pilot supports a limited bilingual live workflow. It does not establish fitness for a professional interpreting assignment.
Questions language professionals may ask
Does a saved source recording become a bilingual transcript?
No. This upload workflow creates text in the spoken language. Translation and bilingual alignment are separate preparation work; the limited live-caption pilot is a different workflow.
Can I rely on the speaker labels?
Use them as editable navigation. Speaker detection can be wrong, so check every attribution that matters against the audio.
How can I deliver the source text?
You can copy it, download a TXT file, or share a revocable, expiring transcript link when the assignment permits that method.
Test the workflow on one recording
Start free: 30 transcription minutes once per account, no card needed - just confirm your email. That is about 15 minutes of audio with automatic speaker separation.
Eligible new customers receive a 7-day, card-required trial with a one-time 120-minute transcription allowance — about 60 audio minutes with automatic speaker separation. After that, the subscription is $8.99/month. Returning customers are charged immediately.
Public checkout is not open yet. Request early access to test a source-language transcription workflow.
Third-party AI providers transcribe audio and Cloudflare R2 stores media. Review the privacy page and the requirements of the assignment before uploading confidential material.