In publica commoda

In support of digital accessibility and research: automatic captioning and transcripts of video and audio recordings

The Amberscript license expired at the end of May 2025. Alternative options can be found below.

Captions and transcripts support digital accessibility: they help hard of hearing students understand spoken language content. For some students, in order to follow the course content, it is necessary to comply with the legal requirement which stipulates that teaching learning materials must be accessible. However, other students benefit as well: Textual alternatives, such as closed captioning and accessible transcripts, are practical for anyone who learns in noisy or noise-sensitive environments, who has a poor Internet connection, or, who learns better visually than auditorily. In addition, they help international students whose first language is not the language of the course.

Automatic subtitles can be generated with free solutions, e.g. via YouTube. However, this approach is not recommended due to concerns from lecturers and data protection issues. The VOICE AI and Melvin services offer a more data protection-safe alternative to support lecturers in the accessible design of their digital teaching and to meet research needs. The services can also be used to make materials for information and counselling services or public relations work more accessible. Choose the service according to your needs.

VOICE AI for transcriptions
(e.g. for research or subsequent editing for accessible transcripts)

For your transcription needs, you can now use VOICE AI (https://voice-ai.academiccloud.de/), which is hosted by KISSKI or the GWDG and is connected to the Academic Cloud. You can use VOICE AI to transcribe video and audio recordings in many different languages in the source language, or to translate the spoken text from the source language into English. The transcribed text can be downloaded as a text file or caption file.
On the GWDG website you will find further information on data protection for VOICE AI.
Note: Please note that the program itself does not provide an editor for the following corrections of the AI-generated texts. You must make corrections using other text editor programs (e.g. copy the text into Word and then format and correct it as required). Please also note that the AI-generated subtitles do not (yet) comply with accessibility standards.

Captions (according to accessibility standards)

If you would like to generate captions for your video material, you are welcome to use the free Melvin service: https://melvin.shuffle-projekt.de/de-DE. You will need to register with a short email, as the service is currently under development. The output results of the captions are very good and the range of functions is similar to that of Amberscript in that there is also an editor within the program and the segmentation or standing times can be adjusted - in addition to other functionalities.
Instructions on how to integrate captions into different systems at the University of Göttingen can be found in the links in the right-hand column.

Quality of speech recognition VOICE AI

The quality of speech recognition is very good thanks to the underlying Whisper model; however, like any automated solution, the reproduction is not perfect and varies according to language, sound quality, accent, etc., so small errors must be corrected. Errors can occur especially with compound words, technical terms, foreign words and punctuation. You can correct errors after downloading the document by copying the text into an editor of your choice. The downloaded transcripts are not yet barrier-free. Please follow our general instructions for creating transcripts and captions in accordance with accessibility standards.