Podcast Workflow

transcriptcut is built around video podcasts: long, multi-speaker recordings that need clean audio, show notes and chapters for publishing, and short vertical clips for promotion. This page walks through that workflow from raw recording to episode + Shorts.

Every AI step below is an explicit button you press. Nothing runs on its own, and AI output never edits your project without you reviewing it (see AI-Assisted Editing).

1. Upload a long episode

Upload the recording from New project. Long episodes are fine:

• Uploads up to 10 GB by default (set MAX_UPLOAD_MB in server/.env to change it). The file streams straight to disk, so size is limited by free disk space, not memory. The upload button shows progress.
• Transcription runs in the background, in chunks. The server extracts a small speech-quality audio track (~14 MB per hour), splits it into ~10-minute pieces at natural pauses so no word is cut in half, transcribes three pieces at a time, and stitches them back onto one timeline. You'll see "Transcribing… 3/7" while it works. An 18-minute episode takes about a minute.
• A 720p editing proxy is made in the background for sources above 720p or over 300 MB, so a multi-GB 4K original doesn't stutter in the browser. Exports always use the original.

2. Edit by editing the transcript

Cut by deleting words, sentences or segments, and use AI tools → Find filler words / Find long pauses for one-click cleanup. See Editing Workflow.

3. Label speakers

Click Detect speakers above the transcript. It labels who said what ("Speaker 1", "Speaker 2", …) and splits segments where the speaker changes. Then, from the speaker chips:

• Click a name to rename that speaker everywhere ("Speaker 1" → "Host").
• Scissors cut everything that person said; the restore button brings it all back.

Names flow into captions, transcript exports and show notes.

Speaker detection is slower than transcription (about 8 minutes for a 14-minute episode), which is why it's a separate button. It stays consistent across the episode even though long recordings are processed in chunks. On a 14-minute synthetic 3-voice test conversation it labeled 99.9% of words correctly. Two very similar voices can end up merged into one speaker; fix those by renaming, or by relabeling individual segments.

4. Make it sound like a podcast

Open the Audio tab and click Make podcast-ready, or pick settings yourself:

Setting	What it does
Loudness	-16 LUFS (podcast apps) or -14 LUFS (YouTube, streaming), with a peak limiter so the louder result doesn't clip
Noise reduction	Light or strong background-noise removal
Level speakers	Evens out a quiet guest and a loud host
Remove rumble	Cuts desk thumps, traffic and AC hum below 80 Hz

The browser can't run these filters live, so Preview 15s from playhead renders a short before / after sample to listen to. The settings are applied when you export. Everything is off by default, so projects that don't use it export exactly as before.

5. Publish

The Publish tab prepares everything around the episode:

• Chapters. Generate them with AI, then rename, remove, or add one at the playhead. Click a time to jump there. Copy for YouTube gives 00:00 Intro lines for a description. Chapters are built into every MP4 and MP3 export, so podcast apps and players show them.
• Show notes. A summary, key points and three title ideas, all editable. Copy all gives Markdown with the chapters appended.
• Transcript. Download the edited episode as TXT or Markdown, with speaker names and timestamps.

Then use the format picker next to Export:

• MP4 video for YouTube and video podcasts.
• MP3 audio (192 kbps, 44.1 kHz) for your podcast feed.
• WAV audio for further mastering elsewhere (WAV can't hold chapters).

Chapters are stored against the original recording, so cuts you make after generating them don't break them: a chapter whose moment you cut moves to the next surviving moment.

6. Make Shorts

The Clips tab turns the episode into short vertical clips:

• Find highlights picks the episode's best standalone moments (15–90 seconds), each with a title and a one-line reason: moments that open with a hook, hold one complete thought, and end when it lands.
• Clip from transcript selection makes your own: select words in the transcript (shift-click for a range), then click the button.
• For each clip, choose 9:16, 1:1 or 16:9, slide the crop position (outlined live on the video), and toggle captions.
• Render clip makes a 1080×1920 (for 9:16) MP4 with big word-by-word captions in your caption font and colors, placed clear of the TikTok / Shorts buttons.

Your edits still apply inside a clip: removed filler words stay removed, and the audio settings apply too. Posting to social platforms is deliberately out of scope. You download the file and post it yourself.

What it costs

transcriptcut itself is free. The AI steps run on your own OpenAI API key, and each one is a button you press:

Step	Model
Transcription	whisper-1 (OpenAI lists it at $0.006 per minute, about $0.36 per hour of audio)
Speaker detection	gpt-4o-transcribe-diarize
Chapters, show notes	gpt-4o-mini
Find highlights	gpt-5.6-luna

Costs for the other steps depend on episode length and current OpenAI pricing. Your OpenAI usage dashboard shows exactly what each run cost. Audio cleanup, rendering, captions and exports run locally with FFmpeg and cost nothing.