Audio & Video Transcription
Unlock the spoken knowledge in your recordings.
Meetings, interviews, and video tutorials hold critical information that traditional search cannot access. Dhito automatically transcribes media files locally, making them searchable by meaning.
The recordings you keep are the files you cannot search
A week of meetings produces hours of audio and almost no searchable text. The decision you need to recall was spoken aloud, attributed to nobody in particular, somewhere inside a file named after the tool that recorded it.
Sending that audio to a transcription service solves the search problem and creates a disclosure problem. Client calls, interviews with named participants, internal reviews, and anything covered by an NDA all become an upload to a third party, subject to whatever that vendor retains and whoever it shares with.
Transcription that never leaves your Mac
Dhito runs OpenAI's Whisper model on your Apple Silicon, using the Neural Engine rather than a remote GPU. Recordings are transcribed in the background as they are indexed, and the audio is never transmitted anywhere. The compliance question that normally accompanies transcription simply does not arise, because there is no transfer to assess.
Each transcript is then embedded into the same semantic index as your documents and images. That means a single query searches across what you wrote and what you said, and it matches on meaning rather than on exact wording — useful, since people rarely speak in the phrasing they later remember.
Results carry timing information, so Dhito returns the moment inside the recording where something was discussed rather than just the file that contains it.
On-Device OpenAI Whisper
Run transcription tasks directly on your Apple Neural Engine. Convert hours of spoken dialogue to text in minutes without uploading files to servers, keeping your meetings completely confidential.
Semantic Speech Indexing
Search transcripts by concept rather than literal words. Type 'budget discussion' to find the exact point where a client discussed project costs in a two-hour recording.
Interactive Video Timestamps
Locating a conversation is only step one. Dhito takes you straight to the exact second a sentence was spoken, letting you open and play the media at that precise timestamp.
Common questions
Is my audio uploaded for transcription?
No. Whisper runs locally on your Mac. Recordings, transcripts, and search queries all stay on the device, and transcription works with no network connection.
How accurate is on-device Whisper?
Whisper handles conversational audio and a wide range of accents well, and running it locally does not reduce its quality. Difficult audio — heavy crosstalk, poor microphones — still produces errors, but semantic indexing means a query usually matches on surrounding context even when a word came through wrong.
Can I search a recording by topic instead of by exact quote?
Yes. Transcripts are indexed by meaning, so searching "budget discussion" can find the point where a client talked about project costs without either of you ever using the word "budget".
Does it identify who was speaking?
Dhito indexes what was said and when it was said, then takes you to that timestamp so you can hear it in context. It does not label speakers by name.
Can I ask questions about a recording once it is transcribed?
Yes. Transcripts feed the same local document chat as your PDFs, so you can ask what was decided in a meeting and get an answer that points back to the moment it was said.