Every word.
Within reach.
Turn conversations, interviews, and voice notes into text you can find your way around.
No speaker count to set. Just upload your recording.00:00 · Speaker 1The little girl had been asleep,
but she heard the raps
and opened the door.
Find it easily.
A transcript that
keeps its context.
No speaker count to set. OpenSEP finds the voices and labels the transcript. Timestamps help you check each turn against the recording.
Try the workspaceNo speaker count to enter. OpenSEP finds the voices automatically.
About this transcript
Four passages read by two narrators, taking turns. Speaker labels and wording come from the recorded OpenSEP engine run. Review important passages against the audio.
Recording sources & changes: Mini LibriSpeech · CC BY 4.0
No speaker count to enter. OpenSEP finds the voices automatically.
Input arrangement · A–EMeet all five voices.
About this transcript
Five real readers in one constructed recording: solo turns, brief interruptions, longer overlaps, then several voices at once. Speaker count was not supplied to the engine. The transcript is its unedited output. This run detected 5 speaker labels. Review wording and speaker boundaries against the recording. Review important passages against the audio.
Recording sources & changes: Mini LibriSpeech · CC BY 4.0
Built around the way you work.
No headcount needed.
Upload your recording without specifying the number of speakers. OpenSEP detects voices and adds speaker labels automatically.
Find the moment you need.
Every line keeps its timestamp, so a moment is easy to find again without scanning an entire conversation.
Take the words with you.
Export a readable transcript or timestamped subtitles for the next step in your workflow.
What should I upload?
Start with a recording where speech is audible. Interviews, podcasts, meetings, and personal voice notes are useful starting points. Only upload recordings you have permission to process.
Do I need to enter the number of speakers?
No. Leave the headcount to OpenSEP. Speaker detection runs automatically; review labels against the recording, especially when voices overlap.
Should I review the transcript?
Yes. Names, technical terms, overlapping speakers, and background noise can affect transcription. Check important passages against your original audio before publishing or sharing.
Can I choose the language?
You can select a language or use automatic detection in the workspace. The output preserves the spoken language; transcription does not translate the recording.
Are my recordings used to train models?
No. Your files are used to process the task you request, then removed on the published retention schedule.
Good ideas deserve to be remembered.
Start with the conversation you already have.