Tech guide
Code Walkthrough Audio: Explain Any Codebase or PR for Async Reviews
From a 600-line diff to a clear spoken walkthrough in one click.
Paste a diff, a PR description, or a chunk of a codebase and walk away with a clear spoken walkthrough your team can listen to instead of squinting at a 600-line diff at 11pm.
A code walkthrough audio is exactly what it sounds like: a spoken explanation of a pull request or a slice of a codebase that a reviewer can listen to before, or instead of, scrolling the raw diff. You paste the change and the AI narrates what it does, why it was made, which files matter most, the risky bits to look at closely, and what the author wants feedback on. Two voices trade off like a senior engineer walking a teammate through the change at a whiteboard.
It works because reviewing code cold is slow and easy to get wrong. A 600-line diff with no context invites nitpicks on formatting while the actual architectural decision sails through unreviewed. An audio walkthrough front-loads the intent so reviewers arrive knowing the shape of the change, and it travels well: an async teammate three time zones away can listen on a walk and leave sharper comments than they would after a rushed late-night skim.
How to make one with Pollinator Studio
- 1
Start from the Code Walkthrough template
In the template gallery, click the pre-built Code Walkthrough template under the Tech category. One click loads the whole recipe: an explainer prompt tuned to summarize intent, call out key files and risky changes, and surface review questions; a clear two-host structure; default voices Alnilam and Puck; a near-silent background bed; and a sensible target length. You can generate right away or adjust any piece first.
- 2
Paste your diff, PR description, or code
Paste the git diff, the PR body, a commit range summary, or a focused chunk of the codebase straight into the content field. Including the PR title and description helps the AI explain the why and not just the what. There is no GitHub or GitLab integration, so copy the unified diff or the relevant files in; for a large PR, paste the description plus only the files that carry the real logic rather than the entire lockfile churn.
- 3
Let AI write the walkthrough script
The template's prompt structures the narration the way a good reviewer thinks: a one-line summary of the change, the motivation, a file-by-file tour of what matters, the parts that need careful eyes, and the open questions the author wants answered. Set your target length first, eight to twelve minutes suits a substantial PR, then review the draft and edit any line directly. Tighten anything the AI over-explained or correct a detail it inferred wrong before you generate audio.
- 4
Assign and tune your two hosts
The template ships with Alnilam as the lead engineer who explains the change and Puck as the curious reviewer who asks the questions a real teammate would. Swap either from all 73 voices and preview them first. Set per-host delivery and pace so the lead stays precise and measured through the technical parts while the reviewer sounds genuinely curious. That back-and-forth is what keeps a dense technical walkthrough listenable instead of a monotone wall of jargon.
- 5
Fix pronunciation and polish the sound
Add pronunciation rules at the workspace or project level for the things AI voices butcher: function names, library names like nginx or PostgreSQL, your service codenames, and acronyms. Keep the background music near-silent or off entirely so nothing competes with the explanation, or pick a low ambient bed from the 83-track licensed library. Generate AI cover art or upload one with your repo name so the walkthrough looks like part of your engineering process.
- 6
Render, then share with your reviewers
Hit generate and the async renderer assembles the final audio in the background while you get on with other work. Download the MP3 and drop it in the PR description, a Slack thread, or your team channel so reviewers can listen before they open the diff. One-click RSS distribution is there if you want a subscribable internal engineering feed on Spotify, Apple Podcasts, or Amazon Music. If this fits your review flow, save it as your own custom template so every future PR walkthrough starts from your exact setup.
Make it your own
The Code Walkthrough template is ready to use as-is — one click and you're generating. But every part is editable: swap any of the 73 AI voices and set each host's delivery and pace, change the background music, edit the AI script and intro/outro prompts, set the length, and add your own or AI-generated cover art. Use the Code Walkthrough template as-is, or tailor every part: swap either host from 73 voices, set each speaker's delivery and pace so the explanation stays clear, edit the AI walkthrough prompt and intro/outro, set the target length to match the size of the change, drop the background music to near-silent for focus, add cover art with your repo or team name, then save it as your own reusable template.
Prefer to start from scratch? Build your own custom template and save your setup to reuse for every future episode.
Tips for a great tech episode
- Always paste the PR description alongside the diff. The AI explains intent far better when it knows why the change was made, not just what lines moved.
- For a big PR, paste only the files that carry real logic and skip generated files, lockfiles, and snapshot updates so the walkthrough spends its time on what reviewers actually need to weigh in on.
- Edit the AI prompt to always end with an explicit 'what I want feedback on' section, so the walkthrough closes by pointing reviewers at the exact decisions you're unsure about.
- Add your stack's tricky names to workspace pronunciation rules once: nginx, PostgreSQL, kubectl, OAuth, and your internal service codenames will be read correctly across every walkthrough.
- Keep the background bed near-silent or off. A code walkthrough is dense enough that any prominent music actively makes it harder to follow.
What you can do with Pollinator Studio
- 100+ ready-made templates — one click to start
- 73 AI voices — preview + per-host delivery & pace
- AI script from a URL, pasted text, or a topic
- 83-track licensed music + transition library
- AI-generated (or upload your own) cover art
- One-click RSS distribution to Spotify, Apple & Amazon
Try the Code Walkthrough template free
30 minutes of audio per month. No credit card, no microphone.
Start freeFrequently asked questions
What should I paste for a code walkthrough?
Paste the git diff, the pull request title and description, a focused set of source files, or a commit-range summary. The more intent you include, the description and the why, the better the AI explains the change rather than just describing the lines that moved.
Does it connect to GitHub or GitLab directly?
No, there's no repo integration. You copy the unified diff, the PR body, or the relevant files into the content field. For large changes, paste the description plus only the files with meaningful logic rather than the entire diff including generated and lockfile churn.
How long should a PR walkthrough be?
Eight to twelve minutes works well for a substantial pull request. Set your target length before generating and the script is sized to fit. For a small focused change, drop it to four or five minutes so reviewers get the context without filler.
Will it pronounce library and function names correctly?
Add them to your workspace or project pronunciation rules once and every walkthrough reads them correctly. This is the fastest fix for things like nginx, PostgreSQL, kubectl, OAuth, and your own service codenames that AI voices otherwise mangle.
Can I keep the walkthrough private to my engineering team?
Yes. Download the MP3 and drop it in the PR description, a Slack thread, or a shared drive. One-click distribution to a private or public RSS feed on Spotify, Apple Podcasts, and Amazon Music is there if you want a subscribable internal engineering feed instead.
