Speech to text transcription software that gets your terms right
Turning speech into words is the easy half. The hard half is getting a surname, a dosage, a case citation or a function name exactly right, then delivering text that already looks like a document. Stenograph focuses on that second half, for live dictation and for recorded files alike.
Listening
Where speech recognition usually goes wrong
Errors in transcription are rarely spread evenly. Ordinary words come through fine, and the mistakes cluster around proper nouns, acronyms, technical terms and anything borrowed from another language. Those are precisely the words a reader notices, and the ones that take longest to correct by hand. A transcript that is almost right still has to be proofread line by line, which erases most of the time you hoped to save.
- Personal and company names written phonetically
- Medical, legal and engineering terms swapped for common words
- Acronyms spelled out or merged into other words
- Foreign phrases forced into English spelling
Custom vocabulary as the accuracy lever
Stenograph lets you tell the recognizer what to expect. Your personal dictionary holds the names and terms you use, written exactly as they should appear, with capitals, hyphens and unusual spellings preserved. On top of it, professional vocabularies for legal, medical, software and finance work cover the standard terminology of each field. Solo includes one professional vocabulary and 500 custom terms, Pro includes all vocabularies and 2,000 terms, and team plans share a dictionary of 10,000 or 50,000 terms across members. The custom vocabulary speech recognition page shows how entries are added and managed.
- Write a term once with the spelling you want
- Add alternatives for how people pronounce it
- Apply the same dictionary to dictation and to uploaded files
Formatting that turns speech into finished text
Spoken language is full of restarts, fillers and run on sentences. Stenograph removes the filler words and false starts, adds punctuation and capitals, and breaks text into paragraphs. When you dictate, it can shape the result as an email with a greeting and sign off, a bulleted list, a memo or structured notes. When you transcribe a file, it separates speakers and keeps a clean reading view alongside the verbatim text. Read more about ai text formatting dictation.
- Fillers and repeated words removed in the readable view
- Lists, numbered steps and headings from natural speech
- Edit by voice with commands such as "make that a list" or "replace X with Y"
Speech to text in more than 50 languages
Many professionals do not speak one language all day. A doctor may quote a term in Latin, a lawyer may cite a foreign decision, a manager may move between English and Spanish in the same call. Stenograph recognizes more than 50 languages and keeps mixed speech as it was spoken, including a switch inside a single sentence. The output stays in the language you used, so nothing is silently translated. See multilingual dictation for the full list and how detection works.
Live dictation and recorded files in one product
Live dictation works at the cursor. You tap once or press a hotkey, speak, and a natural pause ends the utterance; the microphone is on only while you speak. The text appears in the app you are using, through the desktop app for Mac and Windows, the Chrome extension or the phone keyboard. Recorded files go through the upload flow instead, where Stenograph adds speaker labels and timestamps. Both modes use the same dictionary and the same formatting rules, so a term is spelled the same in your email and in your transcript. Corrections you make in one mode improve the other, because a term you add while editing a transcript is available the next time you dictate.
Export the transcript where it belongs
When a transcript is finished, export it as Word for editing, PDF for sharing and records, TXT for plain text systems, or SRT for subtitles with timing. Speaker names and timestamps are included when you want them. Dictated text needs no export at all, because it is already in your document. For bulk work with long recordings and subtitle files, the audio transcription software page covers file handling in detail.
Your audio is not kept
Speech is processed and then deleted, and audio is not stored. Text history is optional and can be turned off or deleted at any moment. Nothing you say or upload is used to train AI models, and all data is encrypted in transit and at rest.
Speech to text questions
How do I improve recognition of a difficult name?
Add it to your dictionary with the exact spelling. If people pronounce it in more than one way, add those pronunciations as alternatives. The next dictation and the next file use it immediately.
Does Stenograph translate what I say?
No. Output stays in the language you spoke. Mixed speech stays mixed, so a sentence that starts in English and ends in German is written that way.
Can I keep the exact verbatim wording?
Yes. File transcripts keep a verbatim view next to the cleaned reading view, so you can quote someone word for word.
Which professional vocabularies are included?
Legal, medical, software and finance. Solo includes one of your choice, Pro and higher plans include all of them.
How are minutes counted?
Dictation minutes count spoken time; transcription minutes count the length of the files you upload. Both are shown in the app each month.
Start with your own terms
Add the twenty words you get wrong most often, then dictate or upload a recording and compare. Pricing begins at 24 USD per month on yearly billing.