A transformer based speech-to-text model for English and German
A new generation of speech recognition
After almost two years of research, we're happy to introduce Ichograph. We build Ichograph to be able to easily transcribe unkown words and beat any other model accuracy for both German and English.
- Class Leading Accuracy
- In our internal benchmark, Ichograph scored best amongst all competing models with only a 2% word error rate (WER)
- Faster than real time
- Our infrastructure creates transcripts significantly faster than real time.
- Build-In Translation
- Sometimes you need transcripts or subtitles in other languages: currently we support generating transcripts in both English and German, regardless of the language of the source material.
- with word level timestamps
We support generating subtitles in various formats, including SRT, VTT, ASS, Word SRT and more.
Additionally Ichograph also can perform speaker diarisation, so you easily create buitiful transcript pages or build apps with word highlighting during playback.