Already have interview transcripts? Start your analysis here

Start with the transcript you already have

The interviews are done. Someone has cleaned up the transcripts. You still need to work out which comments matter, whether different participants describe the same problem, and what you can actually defend in a report.

ScribeFlash accepts an existing transcript as an interview. Choose a UTF-8 .txt file or paste the text. You can then review the source and select evidence without uploading the recording or running transcription again.

Try one interview through the whole process before importing your archive. Combining six months of conversations into one file makes attribution harder later.

Prepare a source you can check, before writing a summary

Keep each interview separate and use a name such as “P01 — first-use interview.” Preserve the order of questions and answers. Use natural speaking turns for line breaks, and retain existing speaker labels or time markers.

This invented example shows a useful format:

Interviewer: Where did you stop when inviting a colleague last time?
P01: I didn't understand what an admin could access, so I asked my manager.
Interviewer: What happened after that?
P01: My manager said it was fine, and I sent the invitation.

Do not rewrite the second line as “The user churned because permissions were unclear.” This participant paused, then continued. That rewrite puts an interpretation into the source itself.

If you only have researcher notes, label them as notes rather than a verbatim transcript. Their summaries can supply context, but should not become participant quotations. For a Word document, copy the body or export UTF-8 plain text. Renaming .docx to .txt will not convert the file.

Review identifying details before uploading

Check that you have permission to use this material for the study and with your chosen tool. Review names, contact details, customer names, order numbers, and combinations of details that could identify someone. Use a stable label: P01 should remain P01 throughout the material.

Keep context that matters to the question. “Schedules shifts at a small logistics company” may be useful; the company's exact name may not be. If you need a lookup table connecting codes to names, store it separately with restricted access.

GOV.UK's data guidance says research data should be used within the consent obtained. ScribeFlash's import checkbox records your confirmation of permission and removal of identifying information. Checking it does not automatically redact the text.

Import a transcript into ScribeFlash, step by step

1. Sign in and create a research project. Set a specific question, such as “What do operations staff need to confirm before inviting colleagues?” Research templates can help define the scope.
2. Open the project's Interviews area, choose Upload interviews, and switch to Existing transcript.
3. Enter the interview name or participant pseudonym. Select a .txt file or paste the body into the text field. You can review the loaded file's text in that field.
4. Scan the content for broken characters, merged questions and answers, and a missing ending after copying.
5. Confirm that you have permission to process the material and have removed personal identifiers, check the confirmation box, and select Import transcript.
6. After import, the source page opens. Check the beginning, end, and a passage in the middle before moving on to the next interview.

If Existing transcript is not available in your interface, contact the team to check availability in your version. The audio/video input is not a place to upload text files.

Check the boundaries of line numbers and speaker labels

Import retains nonempty source lines and their original line numbers. Blank lines are skipped without renumbering the following lines. A gap in line numbers therefore does not necessarily mean missing content.

This import does not infer speakers or recording timestamps. “P01” and “00:15” remain ordinary text when present in the source. Use the interview and source line to verify a quotation; a typed time marker does not provide audio playback.

Correct mistakes carefully, especially negation, amounts, or identifying details. Consult the original material you are authorized to retain. If you change the text after analysis, check the associated citations again.

Begin with one quotation and one research question

In the invented example, select the statement about not understanding admin access and create evidence linked to your research question. Keep your interpretation in a researcher note, such as “Follow up on which permission was unclear.” This keeps the participant's words separate from your next question.

Look for other explanations in the next interview. Someone may understand permissions perfectly but still need approval under company policy. Grouping both accounts as a wording problem could lead the team to fix the wrong thing.

Review supporting and conflicting material before confirming a finding. The research-report guide covers the subsequent handoff. One transcript is enough to practice the workflow; any finding still needs an honest description of its scope.

If import fails, check the file and text limits

Each import accepts up to 120,000 Unicode characters, 2,000 lines, and 12,000 characters per line. TXT files are limited to 480,000 bytes. Character count and file size are different measurements.

For a file error, save a new copy as UTF-8 plain text in an editor that offers an encoding choice, then reopen it to check the contents. If the whole interview is on one line, add breaks at natural speaking turns. When material must be divided, preserve the relationship to the original participant so later counts do not treat the parts as different people.

Each import uses one interview allowance and no transcription minutes. Analysis, reports, and exports still follow your plan's permissions. Successful import does not unlock every later step for free.

References and a useful next step

Preparation references: GOV.UK: Analyse a research session and Managing research data and participant privacy. The cover is an illustrative interview scene, not a customer research record.

Import, inspect, and select evidence from one transcript. Then use the AI finding verification guide to practice returning from a claim to its full source exchange.