Transcript

Audio and video files have a Transcript tab in the file window. This is where a recording becomes a text with time marks and speakers, and where you read, listen to, correct and export it. Recognition runs on the server of your instance, in the transcription service; the recording does not leave the instance for this.

Transcribing

If a recording has no transcript yet, the tab shows the player and below it the settings for the job. The Transcribe button is visible to anyone who may edit the file.

Setting	Options	Effect
Kind of recording	Automatic, Interview (2 speakers), Conversation, Talk (1 speaker), Music, stage	Helps to separate the speakers and to recognise sounds
Speakers	automatic, 1 to 10	A fixed number, if you know it
Language	automatic, German, English, French, Italian, Spanish, Dutch, Polish, Czech, Danish, Swedish	The language of the recording
When	now, in the night window	in the night window starts the job only during the hours set under Rules and night window

After that, the same place shows the state of the job:

State	Display
Waiting in the queue	How many jobs are ahead of it
Waiting for the night window	From what time. Start now starts it straight away, with the same settings
Transcribing	The step (reading the audio, detecting sounds, separating speakers, recognising speech, applying corrections) and the progress in per cent. If the service is busy: waiting for the service with the position
Failed	The reason and Try again

You may close the window in the meantime; the job carries on in the background. If the transcription service does not answer, the tab says so, and the button stays locked until the service runs again.

Audio and video files can also be transcribed without this button: by a rule for a file keyword, see Rules and night window . An upload never waits for a transcription.

The editor

Once there is a transcript, the tab is an editor.

Area	Content
Header	Playback: 5 seconds back, play and pause (also with the space bar), 5 seconds forward (arrow keys), next to it the time, the length of the recording and who is speaking at the moment; if a Loop is running or only one speaker is shown, this appears there with a button to end it. On the right speed and Follow
Timeline	The whole recording in lanes: Wave, Text, Speakers, Language, Sounds (music, singing, applause, laughter, cheering), Doubt and Corrections; plus the clips. Lanes without entries are left out. Above it an overview of the whole recording, which also shows the current position
Manuscript	The text in paragraphs per speaker, each with a time mark. Sounds appear as a line of their own or in the text, and for videos there are stills next to the paragraphs
Sidebar	For videos the picture, then speakers, clips, corrections and the signs in the text
Toolbar	At the bottom: Search the transcript, Undo, Transcribe again, the saving status with the number of corrections, Export, Full screen and How it works (a short help)

During playback the spoken word is highlighted. With Follow, text and timeline follow the current position; if you scroll yourself, it leaves you alone for a few seconds.

Where	What happens
Click on a word	A bar appears above the word: From here plays from this word, next to it are the speaker, the language and how sure the machine was. `Ctrl` + click plays straight away
Click on a time mark	Plays from the paragraph
Click on the timeline	Jumps to that position
Timeline, mouse wheel	Zooms; `Shift` + mouse wheel moves sideways. Zoomed in closely, the text appears in the Text lane
Drag on the timeline or select text	Selects a passage, also across speakers and paragraphs

The search shows the number of matches; Enter jumps to the next one, Shift + Enter to the previous one.

Signs in the text

Sign	Meaning
dotted underline	The machine was unsure
red	Corrected by hand
italic in brackets	Filled in, for example a word in a pause
recognised again	A passage with a set language that the server has recognised again
small flag before a word	Time anchor; red if it was set by hand

The same legend is in the sidebar under Signs in the text.

Correcting

Anyone who may edit the file can correct it. Anyone who may only read it can read, listen, search and export.

Task	How
Change text	Click into the text and type. After a short pause in typing the change is applied and saved; `Enter` or a click elsewhere applies it at once, `Esc` discards it
Insert text in a pause	Select the passage without a word and enter the text
Timing of a word	Drag the time anchor in the text to a different word, or move it in the Text lane
Passage	Select it, then play it, listen to it in a Loop, give it to a speaker, set a language or keep it As clip
Speakers	Click the name, in the text or in the sidebar: rename, set their language, merge with another speaker
Hiding speakers	In the sidebar the eye hides a speaker, the target shows only this one. Hidden parts are collapsed in the text and skipped during playback

Saving is automatic. Undo or Ctrl + Z takes back the last correction, step by step; in the sidebar you can remove each correction individually. The history of the file shows one entry Transcript corrected per session, and the recognition itself as Transcript created (machine). If someone else has corrected the transcript in the meantime, the editor loads their version and tells you so.

Language of a passage. If you set a language for a passage or a speaker, the server automatically recognises these passages once more in that language. This works for languages the recognition knows; dialects such as Low German or Bavarian are noted, and the words stay as the machine heard them.

Transcribe again recognises the whole recording once more, with the same settings as the first time or with different ones. The corrections are kept and also apply to the new result. If the run fails, the previous text stays.

Clips

A clip keeps a passage with a title and tags: select a passage and choose As clip. Clips appear in the timeline and in the sidebar; there you can filter them by tag, play them one by one, edit them (title, tags, start, end) and export them.

Clip export	Format
Audio	MP3
Video	MP4, videos only
Text	`.txt` with title and tags

Several ticked clips can be played one after the other or exported as a compilation; the clips are then joined in time order.

Export

Export in the toolbar at the bottom downloads the transcript with all corrections.

Format	Content
Text (`.txt`)	With speakers and time marks
Subtitles (`.vtt`)	For video players
HTML (Word, MAXQDA)	In the noScribe format, for further editing in Word or MAXQDA

Below the text and in every export are the models used for the transcription, with their licence (CC BY 4.0).