How to Use the Podli Transcript Editor

Once Podli has processed each speaker track individually - AI noise removal, normalisation, and gating - it transcribes the audio and pauses to let you review the result. This is where you make editorial decisions: what stays, what gets cut, and what the final episode sounds like. The final mix, music, and mastering all happen after you finish editing.

Here's how to get the most out of the editor.

How the transcript editor works

The transcript is generated by AI and attributed to each speaker by name. Every word in the transcript corresponds to a precise time range in the underlying audio. When you delete text, Podli marks those segments for removal - deleted words appear struck through so you can see exactly what will be cut before committing. On finalisation, the cuts are applied directly to the individual speaker tracks - before they are mixed together - using a short crossfade at each edit point to keep the result sounding natural.

You're not editing the waveform - you're editing the meaning, and the audio follows.

The editor layout

The editor is split into two panels. The left panel shows the full transcript, broken into speaker segments. The right panel shows AI edit suggestions and this reference guide. Below both panels sits the waveform player, which stays fixed as you scroll through the transcript.

Making cuts in the transcript

Clicking a word seeks the audio to that position, so you can quickly verify the context before making a cut.

Clicking and dragging across words selects a range. Once selected, press Backspace to delete. Deleted words appear struck through - they remain visible so you can review what will be removed and restore individual words if needed.

Right-clicking a word opens a menu with options to delete, restore, edit the word's text, or reassign the speaker from that point onward.

Undo / Redo (Cmd+Z / Ctrl+Z and Cmd+Shift+Z) works across all changes - cuts, speaker assignments, and text edits - so you can experiment freely.

Making cuts on the waveform

The waveform at the bottom of the editor shows the kept audio as a continuous shape, with vertical lines marking each cut point. You can interact with it directly:

• Drag across the waveform to create an audio cut. The corresponding words in the transcript are automatically struck through.
• Click anywhere on the waveform to seek to that position in the audio.

The waveform and transcript stay in sync at all times.

Managing speakers

The transcript is divided into speaker segments, each labelled with a speaker name and timestamp.

Clicking a speaker label opens a dropdown menu where you can:

• Reassign that segment to a different speaker
• Create a new speaker
• Rename the speaker (changes how the name appears everywhere in the transcript)
• Delete the label, which merges that segment into the speaker above

The × button that appears when you hover over a segment header does the same thing as "Delete label" - it merges the segment into the one above.

Hovering between two words reveals a split point. Clicking it creates a new speaker segment starting from that word, which you can then assign to the correct speaker.

AI edit suggestions

The right panel shows automatically detected candidates for removal:

• Filler words - "um", "uh", and other hesitations are highlighted and can be removed in one click. You can restore individual filler words or all of them if you change your mind.
• Long pauses - gaps that exceed typical speaking rhythm are flagged.
• Repeated phrases - false starts and restated sentences are identified.

Clicking a suggestion scrolls the transcript to that word. You can accept the deletion or skip the suggestion to leave it in.

What to cut

Filler and verbal tics

"Um", "uh", "you know", "sort of" - these accumulate over a long interview and slow the pace of the episode. A ten-second delete in text that would take two minutes to locate and trim on a timeline.

False starts and restated sentences

When a speaker starts a thought, abandons it, and restates it cleanly, the transcript makes this immediately visible. Select from the false start to just before the clean version and delete. The crossfade handles the join.

Long silences and extended pauses

Podli's gating step removes most inter-sentence silence automatically. But extended pauses - a guest thinking for several seconds, an unexpected interruption - show up in the transcript as gaps between sentences. Selecting and deleting the blank space between two lines removes the pause.

Off-topic tangents

In a recorded interview, conversations often wander before returning to the main thread. The transcript makes it easy to identify where a tangent starts and ends, select the whole section, and cut it without touching the timeline.

What not to cut

Natural speech rhythm

It's tempting to tighten everything. But removing every pause and breath creates an unnaturally fast pace that listeners find exhausting. Leave natural pauses at the end of complete thoughts - they give the listener time to process what was said.

Skipping the editor

If you're happy with the recording and don't need to make any cuts, you can click straight through to finalise without making any edits. By the time the transcript editor appears, each speaker track has already been individually denoised, normalised, and gated - so in many cases the content is ready to approve and publish as-is.

Finalising your episode

When you're done editing, click Finalise episode. Podli applies all the cuts to the individual speaker tracks, mixes the result together, and masters the output to -16 LUFS.

If you added intro or outro music before processing - either from Podli's royalty-free library or your own uploaded track - it's automatically included in the final mix. The fade durations are set in Processing Options before you start. Podli remembers your music selection and fade settings across sessions, so you don't need to re-enter them each time.

The finished file is ready to upload directly to your podcast host - no further processing needed.

A note on accuracy

AI transcription is highly accurate on clear, well-recorded audio, but it occasionally misreads words - especially proper nouns, technical terms, or heavy accents. You can correct these directly in the editor by right-clicking a word and selecting Edit text. Correcting a transcription error doesn't affect the audio; it only updates the text. The transcript is a tool for navigating your episode, not the final deliverable.