When using Automatic Speech Recognition (ASR) or AI transcription to turn spoken words into timed text, the speech-to-text engine may introduce errors, especially when handling industry-specific terminology, brand names or proper names.
In cases where you need a 100% accurate transcription, you may want to edit the transcript. This article explains the basics of post-editing transcripts.
TABLE OF CONTENTS
- 1. Editing Transcripts
- 3. Minor Edits (Inline) Using the Highlight Confidences Feature
- 4. For Major Edits, Use the Edit Function
- 5. Adjust Timing
- 6. Modifying Speakers (Inline)
- 7. Adding New Speakers (Inline)
- 8. Modifying Speakers across a Transcript
- 9. Mark Completed
As we explained in another article, you can use Limecraft to automatically transcribe your audio or video in minutes. Because it is an automated process, it may be prone to errors. In particular, you will notice that proper names and specific vocabulary may not always be accurately recognised.
While the machine generated transcription might be good enough for certain purposes like indexing or pre-cut, you might need a 100% accurate transcript for compliance reasons, publication or preparing subtitles. In these cases you can review and edit the transcript as suggested below.
Before you Begin
In order to be able to transcribe audio, you must have subscribed to a plan that allows transcription, or in case you are invited a a team member to a workspace, the workspace administrator must have granted you the permission to execute transcription jobs and edit transcripts. (More info)
1. Editing Transcripts
To enable editing, enter the transcription application and switch the transcription application from "VIEW" to "EDIT" mode. Now you can correct the transcription, including segmentation, speakers, punctuation and words.
Note: if you can't toggle between VIEW and EDIT, please consult your workspace admin or account admin to confirm you have the required permissions.
When using Automatic Speech Recognition (ASR) or AI transcription to turn spoken words into timed text, the speech-to-text engine may introduce errors, especially when handling industry-specific terminology, brand names or proper names.
In cases where you need a 100% accurate transcription, you may want to edit the transcript. This article explains the basics of post-editing transcripts.
TABLE OF CONTENTS
- 1. Editing Transcripts
- 3. Minor Edits (Inline) Using the Highlight Confidences Feature
- 4. For Major Edits, Use the Edit Function
- 5. Adjust Timing
- 6. Modifying Speakers (Inline)
- 7. Adding New Speakers (Inline)
- 8. Modifying Speakers across a Transcript
- 9. Mark Completed
As we explained in another article, you can use Limecraft to automatically transcribe your audio or video in minutes. Because it is an automated process, it may be prone to errors. In particular, you will notice that proper names and specific vocabulary may not always be accurately recognised.
While the machine generated transcription might be good enough for certain purposes like indexing or pre-cut, you might need a 100% accurate transcript for compliance reasons, publication or preparing subtitles. In these cases you can review and edit the transcript as suggested below.
Before you Begin
In order to be able to transcribe audio, you must have subscribed to a plan that allows transcription, or in case you are invited a a team member to a workspace, the workspace administrator must have granted you the permission to execute transcription jobs and edit transcripts. (More info)
1. Editing Transcripts
To enable editing, enter the transcription application and switch the transcription application from "VIEW" to "EDIT" mode. Now you can correct the transcription, including segmentation, speakers, punctuation and words.
Note: if you can't toggle between VIEW and EDIT, please consult your workspace admin or account admin to confirm you have the required permissions.
In EDIT mode, you can modify individual words in the editor, or entire paragraphs by using the edit box. In either case, we will keep the corrections in sync with the media, word-for-word, so a text selection in your transcript will always correspond to proper IN- and OUT points in your media. This makes it easier for you to select quotes and fragments later on.
2. Navigating the TranscriptThere are several ways to quickly navigate to a specific area of interest in the Transcript Editor.
- In View mode, simply click on a word, and the player will jump to the corresponding timecode.
- In Edit mode, use Ctrl + cursor (Windows) or Cmd + cursor (Mac) to achieve the same result.
- Alternatively, place the cursor in the desired location and click the Play button in the lower toolbar.
- Use the play/pause buttons in the lower toolbar to speed or slow down the playback speed.
- Use the arrow buttons in the lower toolbar to move forward or backwards in the video.
- Press Shift + arrow to move one frame forward or backwards.
3. Minor Edits (Inline) Using the Highlight Confidences Feature
Start by making small corrections directly in the text. You can fix punctuation, add missing words, or replace incorrect words.
To speed up this process, use the Highlight Confidence feature.
- Click the Highlight Confidence button to highlight words with a low confidence score.
- These highlights indicate where the AI transcriber is uncertain about accuracy, helping you quickly identify and correct potential errors.
4. For Major Edits, Use the Edit Function
In case you need to update or edit larger fragments of text, you can use the Edit Box in the lower toolbar:
- Select the text you want to modify.
- Click the pen icon in the lower toolbar.
- An Edit Box opens, allowing you to safely adjust the text while preserving timecodes and the alignment with the video.
The in and out time of the selection remain intact, while the word-level time codes are interpolated upon confirmation.
Tip: The wider the selection, the higher the risk of misalignment in timing, so best is to keep edits as specific as possible. As a rule of thumb, we recommend to edit sentence per sentence.
5. Adjust Timing
If you notice that the timing of words or paragraphs is incorrect, you can manually adjust the time codes:
- Select the text that requires adjustment.
- Click the clock icon in the lower toolbar.
- An Edit Timing dialog appears, displaying the in and out timecodes.
- Use the arrow buttons to fine-tune the timing or enter the correct values manually.
6. Modifying Speakers (Inline)
Limecraft’s AI transcription automatically detects speaker segmentation, identifying when a new speaker begins and keeping track of which paragraphs belong to which speaker.
By default, speakers are labelled S1, S2, S3, etc., while UU is used when the system is uncertain about the speaker’s gender.
If needed, you can easily edit speaker segmentation and update speaker names to improve accuracy. Here’s how.
- Place the cursor on the speaker's name.
- Remove the existing name and replace it with the correct one.
- A dropdown list will suggest previously saved speaker names for the clip.
For multiple paragraphs spoken by the same speaker, we don't show the speaker name to not disturb the flow of the transcript. To reveal the speaker name, use the F4 or shortcut, or click the coloured line in front of the paragraph
7. Adding New Speakers (Inline)
Depending on the audio quality, it might happen that speaker changes are not picked up by the AI. This is how you can add speakers.
Add Speakers
- Place the cursor where the paragraph should be split and click enter.
- Click to the left of the transcription pane to add a speaker.
- Enter or edit the speaker's name. Removing the speaker number opens a mapping dialogue to select the correct speaker.
8. Modifying Speakers across a Transcript
You can change the name of a given speaker (in all paragraphs spoken by this speaker) using the "Edit Speakers" function.
- Click 'Edit Speakers' in the upper toolbar to open the Edit Speakers dialog.
- Change the speaker’s name and save – all paragraphs with this speaker will be updated.
- Click 'Next Paragraph' to confirm changes for this speaker.
- Use the small grey arrow to move to the next speaker and adjust their name if needed.
More info about editing speakers across transcripts in this article.
9. Mark Completed
Once you have finished editing, it is recommended to Mark the Transcript as Complete. This helps keep track of your progress and allows filters and search functions to identify:
- Clips that still require editing.
- Clips with completed transcriptions.
Additionally, custom workflows can be configured to use the Mark Complete action as a trigger for further processing. By default and without further configuration, "Mark Compete" doesn't have a specific effect. Custom workflows, such as sending a notification, or sending the transcript to a specific destination, must be configured by your workspace admin before becoming effective.
By following these steps, you can efficiently refine transcripts in Limecraft while ensuring accuracy and maintaining proper synchronisation with your video content.