There are several steps you can follow in the process of transcription:

1. Select a target episode. Using the form below, please choose an available episode. Completed or assigned episodes should be automatically filtered out of the list. Some episodes are longer; others are shorter. Select one you feel competent to complete. By submitting which episode you have chosen, it will help us keep track of which episodes are being worked on so we don't double up. Feel free to submit the form and move on to step 2 (without waiting for confirmation from me).

2. Download the audio file to your device. On each episode page, there is a download button to the far right in the “Also listen on” section. To download the file, right click on the button and click “Download linked file” or “Save link as…”, depending on which browser you are using.
3. Transcribe audio with Riverside AI. Upload the audio file to Riverside AI (https://riverside.fm/transcription), wait for the transcription to complete, and then download the transcript TXT file. Note that it may take 5 minutes or more to complete. Some episodes with heavy foreign accents may struggle to complete.
4. Quick manual scan. Open the TXT file, and copy the text into a Google Doc (https://docs.google.com/). You will need a Google account for this. I suggest Google Docs because it will make editing easier. Take a quick scan through the text, replacing any obvious text errors (V Generation, The Generation, incorrect names, etc) with the appropriate text. Only focus on spelling errors at this point. Please remove the start and end buffer text from the episode.
5. Split your raw text into chunks. Try to split your text into chunks ranging between 1,000 and 1,500 words per chunk. You will want to use the Word Count feature in Google Docs to do this properly (Tools > Word Count). This will be important for the next step. Try to create a line break between each chunk for tracking purposes.
6. Process the text with Chat GPT. You will need a free account at chat.openai.com. Once you’ve set up your account, there is a prompt I’ve developed that should instruct the AI on how to edit the text. Here is the prompt:
   

   You are an editor and transcriptionist. I am going to supply you in the next prompt with a transcript taken from the audio of a Podcast episode. When I do, please do the following:

   1. Please check for grammatical errors and punctuation issues. Make corrections as necessary only for the sake of proper grammar and punctuation.

   2. Please split the text into appropriate paragraph divisions.

   3. Do not remove, skip, or summarize any sentences from the text! Only edit or remove what is necessary for the sake of grammar. This is very important.

   Once you have made the corrections, please reply with your edits formatted in a way I can use for an article form of the podcast. Please confirm that you understand the instructions and are ready to receive the transcript.

   Once you have copied and pasted the above prompt into ChatGPT, and clicked the “Send message” button, ChatGPT will tell you it is ready to receive the text.

   Now, you will not be able to paste the entire episode in one shot, or it will not fully process the text, which is why we had you split the text into chunks. Take your first chunk of text, paste it into the message field, and click “Send message.” Once you submit the text, the AI will do most of the heavy editing for you and will begin to return the edited text.

   Note that you may need to click “Continue Generating.” Also note that if you enter too much text, ChatGPT may not finish the entire job, which is why I suggest limiting the chunk to 1000-1500 words. Make sure the end of ChatGPT’s reply is the same as the end of the text you entered, and do a quick scan to make sure it didn't skip any large sections. Generative AI can be a little finicky about this from time to time. If went wrong and it ended early or if it skipped a large section, you will need to start a new conversation, enter the original prompt and try again will a smaller chunk of text.

   Repeat this process in a new conversation for each remaining chunk until ChatGPT has processed all of the transcribed text.

7. Reassemble text. Take all of the chunks that ChatGPT produced and copy/paste them into your Google Doc.

8. Listen through and final editing. This step will likely take the longest. Listen through the podcast episode with the edited text open and in front of you. There are a few things you will want to look for in this step:

   Check for accuracy. As you listen, make sure the editing text is basically close to the original. Slight alterations in wording is acceptable, but if the transcription is wrong or clearly confused, make the necessary changes. Don’t worry about correcting the grammar, as sometimes the text is following what was actually said, which may not follow perfect grammar.

   Check for gaps. Every one in a while, you will discover that ChatGPT or the Riverside AI missed a small section of text. If this is the case, please manually add the text to the document. You will probably need to pause and rewind the episode to do this properly.

   Add speaker labels. To do this, add the speaker’s full name in bold at the beginning of the paragraph with a “:” after their name (e.g., Bobby Bosler:). If the episode is a monologue (one person speaking), you only need to add one speaker label at the very beginning of the episode. If the episode is an interview (more than one speaker), please add a speaker label each time the speaker changes. Include the speaker's full name for their first interaction, and then only use their first name for every subsequent interaction. See the Gillmores to PNG episode transcript for an example.

   Other formatting considerations. We have tried to include quotation marks if the speaker is clearly quoting someone. If different paragraph breaks make more sense to you, feel free to create different paragraphs than ChatGPT created. Try not to allow for paragraphs that are too long, as that increases reading difficulty.

9. Transcript submission. Once you have completed the transcription and editing, please paste your transcript into the form below, select the episode that is assigned to you, and submit the form. Once you have emailed the completed transcript, your job is complete!

As soon as we receive the transcript and double check for quality, we will post the transcript to the episode. If you have any questions, please feel free to send an email to bobby@theegeneration.org.