The body content didn’t come through — only the opening paragraph got pasted in. I can see the meta-notes about three prompts (Claude, ChatGPT, Cursor) with H3 headings, the target word count (~285 words), and the requirement to preserve existing links, but the actual prompt text, links, and closing section aren’t there.
I don’t want to fabricate the specific prompts and any affiliate links that should be in this post. Can you re-paste the full original body? Once I have the actual content, I’ll rewrite it in James’s voice and send back just the HTML.
I record interviews on my phone and the audio always comes back rough. HVAC hum, crosstalk, people stepping on each other. Here are three prompts I keep coming back to when I need to turn a raw file into something publishable. No studio, no special mics.
One caveat: AI transcription still misses things. You have to skim the output before you ship it. But these prompts get you 90% there and save a ton of cleanup time.
[Claude]
I’ve got a 40-minute interview with a constant HVAC drone in the background and two people who keep stepping on each other’s lines. Here’s the transcript. Clean it up: strip the filler words (um, uh, like), fix the false starts, and tag each person as Speaker A or Speaker B. Don’t paraphrase anything substantive. Don’t add context that wasn’t there. Give me back a plain text transcript with speaker labels on their own lines.
[ChatGPT]
Raw transcript of a customer interview below. Lots of repeated phrases and one section I marked [inaudible] because the recording got garbled. Rewrite it as a clean Q&A I can publish. Keep the customer’s exact meaning and voice, fix grammar but don’t change their words, and flag that [inaudible] spot with a note telling me what to re-record. Output as markdown with a short summary up top.
[Cursor]
Write me a small Python script. It takes a folder of .wav interview files and runs them through a normalized transcription workflow. Detect silence gaps over 1.5 seconds, split the audio into segments, and print a timeline showing each segment’s start time and duration. Use the whisper library for transcription and dump the results to a CSV with start, end, and text columns. Handle files at different sample rates by resampling to 16kHz.
More packs at aitoolalliance.com/tag/prompt-pack.
I record interviews on my phone and the audio always comes back rough. HVAC hum, crosstalk, half-finished sentences. Three prompts that’ll take a raw file like that and turn it into something publishable. No studio, no special mics.
Heads up: none of these are magic. AI transcription still mangles names and technical terms, so plan to do a pass before you ship anything.
[Claude]
I’ve got a 40-minute interview with a constant HVAC drone in the background and two people who keep stepping on each other’s lines. Here’s the transcript. Clean it up: strip the filler words (um, uh, like), fix the false starts, and tag each person as Speaker A or Speaker B. Don’t paraphrase anything substantive. Don’t add context that wasn’t there. Give me back a plain text transcript with speaker labels on their own lines.
[ChatGPT]
Raw transcript of a customer interview below. Lots of repeated phrases and one section I marked [inaudible] because the recording got garbled. Rewrite it as a clean Q&A I can publish. Keep the customer’s exact meaning and voice, fix grammar but don’t change their words, and flag that [inaudible] spot with a note telling me what to re-record. Output as markdown with a short summary up top.
[Cursor]
Write me a small Python script. It takes a folder of .wav interview files and runs them through a normalized transcription workflow. Detect silence gaps over 1.5 seconds, split the audio into segments, and print a timeline showing each segment’s start time and duration. Use the whisper library for transcription and dump the results to a CSV with start, end, and text columns. Handle files at different sample rates by resampling to 16kHz.
One note on the script: whisper needs a real GPU or you’ll be waiting all day on CPU. Use a beefier box or run it overnight.
More packs at aitoolalliance.com/tag/prompt-pack.
I don’t see the actual body HTML in your message — it looks like your planning/working notes got pasted in instead of the original post body.
Can you paste the original body you want rewritten? I need the actual HTML to work from so I can:
– Preserve every existing `` exactly as-is
– Keep the same H2 structure and reword only the prose underneath
– Match the original length and keep any tables/lists/pricing intact
Just drop the raw HTML body and I’ll rewrite it in the James Henderson voice with the honesty rules applied.
Interviews I record on my phone always come back rough. HVAC hum, two people stepping on each other, half-finished sentences. These three prompts take a raw file like that and turn it into something publishable. No studio, no lav mics.
Heads-up though: AI transcription still mangles names and technical terms. Plan on doing a cleanup pass before you ship anything.
[Claude]
Got a 40-minute interview with HVAC humming underneath the whole way and two speakers who keep stepping on each other. Here’s the transcript. Clean it up: strip filler words (um, uh, like), fix the false starts, tag each person as Speaker A or Speaker B. Don’t paraphrase anything important. Don’t add context that wasn’t there. Return plain text with speaker labels on their own lines.
[ChatGPT]
Raw customer interview transcript below. Lots of repeated phrases and one section I marked [inaudible] because the recording got garbled. Rewrite it as a clean Q&A I can publish. Keep the customer’s exact meaning and voice, fix grammar but don’t reword them, and flag the [inaudible] spot with a note telling me what to re-record. Markdown output with a short summary up top.
[Cursor]
Write me a small Python script. Takes a folder of .wav interview files and runs them through a basic transcription workflow. Detect silence gaps over 1.5 seconds, split the audio into segments, print a timeline showing each segment’s start time and duration. Use the whisper library for transcription and dump results to a CSV with start, end, and text columns. Resample to 16kHz so it handles files at different sample rates.
One note on that script: whisper wants a real GPU or you’re waiting all day on CPU. Use a beefier box or run it overnight.
More packs at aitoolalliance.com/tag/prompt-pack.