This is the exact process I use to turn a raw recording into a finished audio-only podcast with Astra and Descript. You bring the conversation. Astra handles the edit with your direction. The setup, the starting prompt, and my full editing playbook are all on this page. Start with one episode.
Everything you need to run this is on this page, one worker inside a broader AI operating system for your business. Four steps gets your first episode into Astra's hands.
Go to ChatGPT → Plugins → search "Descript" → Install Plugin. Connect your Descript account if prompted, then start a new chat.
Descript is where the podcast edit happens. Astra carries out your instructions using the plugin tools or, when available, screen control. If you're using screen control, have the Descript desktop app installed and signed in.
Choose Work → GPT-6 Astra → High.
For files and apps on your computer, use the desktop app with Work locally selected. Make sure Astra can access your recording and your music. If a local folder isn't accessible, attach the files directly. The options in your account may differ, so use the model picker and plugin prompts to confirm your access.
Attach your raw recording. It can be audio or video. The finished episode will be audio only. Include:
Label the files clearly so Astra knows which is which. A short pause before you begin the show makes the edit easier. You don't need a specific opening phrase.
Paste the starting prompt from the next section into the same message as your attached files.
Add your episode title, which file is which, and any opening or closing instructions you already know. Ask for the finished MP3, the editable Descript project, and short opening and closing previews.
Copy this exactly. Paste it into the chat with your recording and music attached.
I want to edit an audio only podcast. I want you to use Descript (which you can use my screen or the plugin). The file is attached. For this, the edit should be simple, create a clean start and clean end, cutting out the dead air. Add in my intro music (attached).
Put the attached pre-roll at the very beginning, in full. Start the show music under my spoken opening after the pre-roll.
For this workflow, cutting dead air means cleaning up the beginning and ending. The conversation in the middle stays intact. If you aren't using music, remove the music sentence and tell Astra to skip it.
This is a simple production edit. It doesn't automatically mean removing every pause, filler word, or tangent. It means a deliberate opening and a complete ending.
The opening should feel intentional. Astra finds the moment you shift from setup or casual talk into addressing the audience, then preserves your complete opening thought and a natural breath before it.
There is no universal timestamp or magic phrase. "I'm excited" was specific to one recording. "What's good" might be a clue in my episodes, and the surrounding context is what decides where the show actually starts.
The ending should feel complete. Keep the entire final thought and sign-off, then remove the off-air conversation and the trailing silence.
Allow a brief natural decay after the last word. Never cut off a syllable, and never stretch the episode just to fit the music.
The order is simple, and it's the same every episode once you lock it in.
Pre-roll, if supplied → spoken opening with music underneath → conversation → final five seconds with music fading out.
Keep the pre-roll intact. Start the show music after it, underneath your spoken opening. Use a short opening music bed, then fade it away so the conversation continues naturally.
Bring the same music back under the final five seconds at the same approved volume, then fade it out as the episode ends.
The music should be clearly audible while every word stays easy to understand.
There isn't one correct dB setting for every music file. In my edit the supplied music was already a quieter version, so we used 0 dB clip gain, which means leaving that file's original volume unchanged. Reducing it by another 8 dB made it too quiet.
That's a reference for that one file, not a preset for every podcast. Ask Astra to balance your music against your voice, then listen to the exported preview. Once you approve the balance, save the setting with that exact music file for future episodes.
If the music is missing, ask Astra to check that it is on the timeline, unmuted, and included in the audio export before changing the volume.
These are the rails that keep the process from butchering your episode. Build them in from your first edit.
The opening lives in this recording, not the last one. Astra inspects this episode fresh, finds the real start, and confirms the music is actually in the exported file. Never accept "done" without a preview you can hear.
The music sitting right under your voice is the difference between amateur and finished. One exported preview, your ears, your approval. Lock the setting to that exact music file and reuse it every episode.
A finished episode that sounds like you, with a clean opening, a complete ending, and music in the right places.
Working with video from the same recording? Run the same process for editing a video.
Ship one episode. Then make the next one faster.
Undeniable Studio is a live building room where you co-create with AI every Tuesday and leave with something real. This is one of the systems we build together.
Join The Studio →