All podcasts and audio-only files at UC Berkeley must have accurate and edited transcripts, regardless of the platform they're hosted on.
- Make sure your transcript is available on every platform that your podcast is hosted on.
- Be accurate. Written transcripts need to accurately reflect the audio content. Do not paraphrase.
- Identify speakers each time they speak. This is crucial if there are multiple speakers.
- Describe important background sounds. Only include background sounds if they’re important to the context or meaning of the video.
Example: [music], [laughter], [audience applause], [dog barking]
- Use correct spelling, punctuation and capitalization.
- Make sure transcripts are accessible. Refer to document accessibility guidance if your transcripts are Google Docs, PDFs or Word files.
- Don’t rely entirely on auto-transcription. Auto-transcripts are a great way to get started, but they will not be accurate without additional editing.
- Don't write your transcript in all caps. Only use all caps for YELLING or speaker identification.
- Don't include timestamps. You want your transcript to be easily read without additional clutter.
How do I build an accurate transcript?
You can choose to pay for a professional transcript, or you can build your own transcript.
Professional transcription services
- 3Play Media: The UC System has a contract with 3Play Media. Learn how to submit a purchase order.
- Convert your audio file to a video file.
- Upload the video file to YouTube and allow YouTube to auto-generate captions for your video. It may take up to 24 hours. NOTE: You can set your YouTube video to "Private" if you do not want it to be publicly viewable.
- Edit your auto-generated captions so they're are accurate and grammatically correct. Add speaker identification. Learn more about How to Edit Auto Captions in YouTube.
- Once your captions are edited, go to the video viewing page. Display the transcript by clicking the 3-dot icon below the bottom right corner of the video. Copy and paste the transcript text into a document.
- Audioship: If you don’t have a way to convert your audio to video (for YouTube auto captioning), Audioship will convert your file and place it on your YouTube channel for you, where it will be auto captioned.
- Use YouTube’s caption editing tool to fix the captions, and then download the transcript.
- Next, you’ll have two options:
- Delete the video.
- Share the podcast as a video with closed captions as well as a transcript.
- Google Docs Tool: Voice Typing: If you have clear, high quality audio, it is possible to use this tool to transcribe the audio into text in a Google Document. It will require clean up, but may save you time typing.
- New AI transcription tools (coming soon)
Best practices for accessible transcripts:
Accessible, well designed transcripts are easy to read and understand. The output will be similar to the script for a play.
- Accuracy: Never paraphrase. Don't remove swear words unless bleeped out; if so, indicate this with [bleep]. You may remove um, ah, mm-hmm, and false starts to improve clarity. If you're unsure, or if they add meaning or context, include the filler words.
- Slang: Include words as they are spoken, as long as readability is not compromised. Examples:
- Identify each time the speaker changes. Use spacing, bold, and/or ALL CAPS to make this clear. If the speaker’s name is unknown, some alternatives are: Student, Speaker #1, Professor, Audience member, etc. If there are multiple unknown speakers use numbers. Example:
STUDENT #1: Hello.
STUDENT #2: Good morning.
- Use paragraphs and spacing to make the text easy to read. Avoid the dreaded wall of words.
- Music: Describe the background music style if needed, in brackets.
- Punctuation: Use punctuation to indicate the speed or pace of a sound effect. You can use an ellipsis for extended pauses, commas for brief breaks, and dashes for quick repetition.
Example: Oh... my... g-g-god. Oh, Em, Gee.
- Speaker Tone: It may sometimes be appropriate to add a description of the speaker’s tone in brackets.
Example: [whisper], [aggravated]
- Math: If transcribing math content, use only numerals. For all other topics, write out numbers 1-10 (one, two), and use numerals for numbers over this (11, 53, 978), or use a combination for large, rounded numbers (3 million).
Examples of good transcripts
- Berkeley Talks transcript: Emiliana Simon-Thomas on where happiness comes from (revisiting)
- Freakonomics transcript: The Economics of Everyday Things: “My Sharona”
- This American Life transcript: Three Miles
- Work Appropriate transcript: Remote Work Done Right with Marissa Goldberg
Who uses transcripts?
Transcripts are crucial for people who are Deaf/Blind. In the absence of captions, the following users may also rely on transcripts:
- Hard of hearing
- Have auditory processing disorders
Transcripts are also useful for:
- People who want to skim the content quickly
- Visual learners
- Users who want to improve retention of information
- Search engines
What’s the difference between captions and transcripts?
- Captions are displayed on screen in short snippets and synchronized with the audio. Captions are typically associated with video content.
Transcripts are the text output of an entire audio recording. Often, they’re provided on the same web page as the media player that contains the audio (or video) file. It may also be provided as a separate document. Transcripts are typically associated with podcast and video content.