SAYRIL is a voice typing program for Windows and Linux: you speak, and the text appears in the window where the cursor is. Speech recognition, translation into 44 languages and text editing run on your computer; SAYRIL does not send your voice to the internet.
Dictation
Three ways to dictate at the cursor
Place the cursor, hold the key and speak. When you stop, the text is inserted where the cursor was. If the focus has moved, the text goes to the window where you started dictating.
Dictation. Hold Ctrl and speak: when you stop, the text is inserted where the cursor is
Dictate and send. Hold the left Shift: after inserting the text, the program presses Enter. Handy in chats and AI assistants
Without holding a key. Ctrl+Space starts and stops recording
Auto-stop after 2 seconds of silence, recording up to 120 seconds (both values can be changed)
Text cannot be inserted into windows of programs run as administrator: it stays on the clipboard
Listening0:02
Hello! The October report is attached.
All features
Dictation at the cursor
Place the cursor, hold the key and speak. The text goes into an editor, browser, email, messenger or AI assistant chat. The previous clipboard contents are restored. Text cannot be inserted into programs run as administrator: it stays on the clipboard.
DictationIn all plans
Dictate and send
Hold the left Shift key: after inserting the text, the program presses Enter, so the message is sent automatically.
DictationIn all plans
Speech translation
Speak in one language, and the text is inserted in another. Translation into 44 languages.
TranslationIn all plans
Translating selected text
Select text in another program and press the key: SAYRIL translates it and replaces the selection. Up to 4,000 characters.
TranslationIn all plans
Audio and video transcription
Transcribes files with up to 3 hours of audio to TXT, SRT and VTT, with subtitle translation. Tested with MP3, M4A, WMA, MP4 and WAV.
Files and meetingsIn Pro and Corporate Pro
Meeting recording and speaker separation
Microphone and computer audio in one file, up to 4 hours. The recording banner is visible the whole time. In the transcript, lines are labeled “Speaker 1, 2…”, up to 8 speakers.
Files and meetingsIn Pro and Corporate Pro
Cleanup, dictionary and replacements
Punctuation and capitalization, filler words, a replacement dictionary for terms and service names. A filter for phrases made up during silence.
Text qualityCleanup in all plans, dictionary in all except Base
Text editing by the model
Fix mistakes, make it formal, shorten: a language model edits the text. Editing is off by default; the model can change the meaning, and the program warns you about this.
Text qualityIn Pro and Corporate Pro
Program profiles and history
Each program can have its own language, translation and insertion method. Dictation history stays only on your computer, is kept for 30 days by default and can be turned off.
Profiles and historyHistory in all plans, profiles in Pro and Corporate Pro
Voice commands
Your own phrases for snippets and keys: insert a template with the date, press Enter or Tab, undo the last insertion.
Profiles and historyIn Pro and Corporate Pro
Graphics card or processor
The program chooses the graphics card or the processor on its own and picks the recognition model by language. Only the regular graphics driver is needed.
HardwareIn all plans
For small teams
Shared organization dictionary, a recognition server on the local network (off by default, unencrypted connection), silent installation with a key.
HardwareServer in Corporate and Corporate Pro, shared dictionary in all plans except Base
Text quality
Cleanup, dictionary and replacements
Spaces, basic punctuation, a capital letter at the start of a sentence
Filler words such as “uh”, “um”, “mmm” are removed, and a strict mode can be turned on if you want (Russian and English)
Replacement dictionary “as recognized → as written”: computer terms, service names, abbreviations
Filter for phrases made up during silence. It is not guaranteed to catch them all: when in doubt, the text is kept
Text editing by the model
In Pro and Corporate Pro: “Fix mistakes and punctuation”, “Make it formal”, “Shorten” and up to 5 modes of your own. The model is off by default. It may change the meaning, and the program warns you about this. If the model does not finish in time, the original text is inserted.
Profiles, history and commands
Program profiles
Pick a program and set its own language, translation, insertion method and whether to press Enter after inserting.
Dictation history
Only on your computer: time, text, language and program, without audio. 30 days by default; you can choose 7 or 90 days or turn it off.
Voice commands
Up to 200 phrases of your own: insert a template with the date, press Enter or Tab, undo the last insertion.
Graphics card or processor
The program chooses the graphics card or the processor on its own and picks the recognition model by language. Only the regular graphics driver is needed. The recommended specs page lists the graphics cards that speed things up.