Transcribe and translate

When importing a video or audio file into a project, we’ll automatically detect the spoken language, generate a transcript and create a overview summary of content in your file. Transcripts can also be translated to one of 75 available languages to streamline collaboration across global teams.

At this time, Dovetail cannot automatically detect every language that we currently support. If the spoken language cannot be detected, a transcript will still be generated; however, it will not be accurate.

Due to this, you can set a default language for transcription at both the project and workspace level.

• To set a default transcription language for a project, open the magic automation settings by selecting the magic icon in the top right corner of the page.

• From there, open the drop-down menu next to Language to select your language. Once chosen, this transcription language will automatically apply to all new transcripts generated in your project.

Workspace admins can set a default transcription language for your entire workspace, while managers or contributors can set a default language for individual projects. This is great if your video and audio files are always in the same language.

• To set the default workspace transcription language, workspace admins will need to navigate to Settings → Transcription → Workspace transcription language.

You can change the language of an existing transcipt at any time.

• To do this, open the note with the incorrect transcription and navigate to Transcription options (•••).

• Next, click Language, select the correct language from the dropdown menu and press Regenerate.

• From there, you can also choose to save this language for future uploads to this project, which will set the project’s default transcription language.

You can transcribe conversations in 41 languages and receive a full native-language transcript back. Please find a full list of supported languages below.

Unsupported languages for auto-detect

Unfortunately, Dovetail cannot automatically detect every language that we currently support at this time. If the spoken language cannot be detected, a transcript will still be generated; however, it will not be accurate.

Please refer to the list of unsupported languages below before using auto-detect.

While our AI speech-engine will attempt to automatically detect different speakers, there are times when it doesn’t get it right.

You can merge two speaker sections into one by placing your cursor at the start of the second section and pressing backspace.

Similarly, you can split one speaker section into two by placing your cursor inside and pressing enter.

You can assign an existing contact, create a new contact or add a placeholder name to a single speaker across an entire transcript. For example, when updating the name of Participant 1, this will be applied to every section this speaker has been identified as within the transcript.

• To add a name to a participant, click on a Participant (Participant 1, Participant 2 etc.)

• From there, enter and select the name of an existing contact, create a new contact, or add a placeholder name for the speaker in the transcript.

For when you need human-level accuracy with your transcripts or to analyze conversations in a language that we don't yet support – we've also added the option to bring your own transcript (in the form of a caption file).

We support importing any WebVTT (Web Video Text Tracks) caption file. When you upload a .vtt caption file, we'll use the caption timestamps to sync with your video or audio file playback.

• To import your own transcript, click ••• below a video or audio file, select Upload transcript, then choose a compatible .vtt file.

To improve the accuracy of transcripts, you can submit a list of custom words or phrases that are not found in a dictionary (for example, company names or industry jargon) before starting a transcription. You can apply custom vocabulary at both the project and workspace level.

In a single project

• To do this, upload your file, select Begin, and navigate to Custom vocabulary.

• From there, enter words or phrases into the textbox and select Start transcription when complete.

In all projects across a workspace

To avoid entering terms every time before a transcription, workspace admins can save a global list of up to 100 words or phrases in Transcription settings under Workspace custom vocabulary. This will apply to all transcripts users create across all projects.

You can now translate an entire transcript and summary to another language to simplify knowledge sharing across global teams. Workspace admins on premium plans can enable the translation beta by navigating to Settings → Beta.

• Once enabled, open the video or audio you want to translate.

• Users will find the option to Translate data under the meatball menu ••• in the top right.

• After selecting a target language, a new copy of the data will be created in the target language, where it can be summarized and highlighted without affecting the original data.

You can translate data in 75 languages and receive a full native-language transcript back. Please find a full list of supported languages below.

Can I download a transcript from my project?

Yes! Check out Download project data with information on how to do this.

How long should file import and transcription take?

Your recording will be uploaded into a note and processed to ensure fast playback. The amount of time this takes depends on the length of the recording. In general, processing takes about 30% of the length of the file; e.g. a 60 minute recording will take approximately 20 minutes to process and transcribe.

You can close the note and continue using other parts of Dovetail while your file is uploading. You can safely leave Dovetail entirely (e.g. close the browser window or turn off your computer) while it’s processing and being transcribed, and come back later as well.