What transcript formats does BrassTranscripts support?

BrassTranscripts delivers transcripts in TXT (plain text), JSON (structured data with speaker labels and timestamps), SRT (subtitle format), and VTT (web subtitle format). For bulk accounts, transcripts can be formatted as flowing paragraphs — one paragraph per speaker turn, labeled once at the top, with no inline timestamps. This format was built after a qualitative researcher told us the standard segmented format broke her analysis workflow. Contact support if you need a specific output format for your work.

Can I rename speaker labels in BrassTranscripts transcripts?

BrassTranscripts labels speakers automatically as SPEAKER_00, SPEAKER_01, etc. These labels appear consistently throughout the transcript. For renaming, download the TXT or JSON file and use find-and-replace in any text editor or word processor to substitute real names. Most qualitative researchers handle this step themselves after downloading, since only they know which voice belongs to which participant.

A Healthcare Researcher's Request for Paragraph Format

A health-services researcher came to us last year with 39 qualitative interview recordings — clinical staff from community health centres, each 25-60 minutes long. Nurses, pharmacists, nurse practitioners. She uploaded them in a single batch and paid when they finished.

Then she looked at the transcripts.

The format wasn't what she needed.

What the Standard Format Looks Like

Our default transcript output segments the audio into timestamped turns. Each time a speaker pauses or changes, a new segment appears:

[00:00:04] SPEAKER_00: The main challenge we've been seeing is 
[00:00:08] SPEAKER_00: the wait times for referrals, which can 
[00:00:12] SPEAKER_00: run three to four months at minimum.
[00:00:16] SPEAKER_01: And is that consistent across all the sites
[00:00:19] SPEAKER_01: you're working with, or more specific to —

For content creators, legal teams, and anyone who needs to navigate audio by timestamp, this works well. For qualitative researchers doing thematic analysis, it's a problem.

Qualitative analysis reads across a speaker's full statement — not segment by segment. Seeing one continuous thought broken into five timestamped lines makes coding and annotation slower, not faster.

What She Asked For

She didn't file a support ticket. She sent a direct email: the format was unusable for her analysis workflow, and could we do something about it.

She explained what she actually needed: one paragraph per speaker turn, the speaker labeled once at the top of each turn, no timestamps in the body. Like this:

SPEAKER_00
The main challenge we've been seeing is the wait times for referrals, which can run three to four months at minimum. And that's just to get the initial appointment — the actual treatment timeline is separate.

SPEAKER_01
And is that consistent across all the sites you're working with, or more specific to the urban centres?

She would rename SPEAKER_00 and SPEAKER_01 to real participant identifiers herself — she knew which voice was which, and pseudonymization was part of her research protocol.

What We Built

BrassTranscripts processes audio and stores the raw transcript data as structured JSON — every segment, every speaker label, every timestamp. The format the customer downloads is generated from that stored data at download time.

That meant we could add a new output format without reprocessing anything. We built a paragraphs formatter: it merges consecutive segments from the same speaker into a single paragraph, adds the speaker label as a header, and strips inline timestamps from the body. The raw JSON stays intact; the download changes.

Her account was set to the new format. Every future download — including re-downloads of her already-processed files — would come out as paragraphs.

She re-downloaded her completed transcripts. They came out exactly the way she described.

What Made This Work

The format request was specific and actionable. She knew what she needed, knew why the default didn't fit, and communicated both clearly. That gave us something to build toward rather than a vague complaint to manage.

The technical path was also unusually clean. Because transcripts are generated from stored JSON rather than saved as static files, adding a new output option didn't require reprocessing 39 audio files. It was a formatting function, not a pipeline change.

But the more durable lesson was simpler: she knew her workflow better than we did. Qualitative researchers read transcripts differently than attorneys or podcasters. A format that works for one use case actively hinders another.

The Service Principle

Building software means making assumptions about how it will be used. Those assumptions are wrong in ways you can't anticipate until someone who uses the product differently tells you.

This researcher told us. We changed it.

The question "how do you actually use this?" is worth asking early and often. The answer is usually more specific — and more useful — than anything you'd design for in advance.

BrassTranscripts now supports paragraph-format transcripts for bulk accounts. If your analysis workflow works better with a different output structure, contact support and describe what you need.

AI Prompts for Qualitative Researchers

Once you have a clean paragraph-format transcript, AI tools can accelerate the analysis work that follows. Two prompts from our guide are particularly useful for interview research:

Qualitative Research Thematic Analysis — Systematic thematic coding of an interview transcript. The prompt follows standard qualitative coding principles: identify recurring themes, group them into categories, and return a structured summary with supporting quotes from the transcript.

Clean Text Extractor — If you need to strip remaining speaker labels and timestamps to produce plain readable prose, this prompt converts a timestamped transcript into flowing paragraph text with no labels or markers.

Both prompts are copy-ready and work with ChatGPT, Claude, or any AI tool.