Podcast voice: make a clear story before buying a microphone
Practise radio and podcast voice work: write a script, record a voice test and prepare an episode with a transcript. Use the running order and exercises to review your work. This free guide has no enrolment or certification.
A piece with clear limits
Five-minute episode: question 20 s; context 40 s; interview 120 s; source comparison 80 s; close 40 s. The total is 300 seconds. At 150 words/minute, 600 scripted words take 240 seconds, leaving 60 seconds for clips and pauses. Time the real delivery.
Write for a listener
Decide who needs the episode and what they should understand at the end. Shorten sentences, attribute claims and mark unfamiliar names for pronunciation. Read the script aloud. When a sentence requires a breath in the middle of its main idea, rewrite it. Open with a specific question rather than a promise the episode cannot keep.
Record a real test
Use a quiet room and a consistent microphone distance. Record thirty seconds of speech and a little room tone, then listen for echo, hum and clothing noise on ordinary headphones. Change the room or microphone position before reaching for processing. Editing cannot reliably restore a word that was never intelligible.
Edit without changing meaning
Remove repetition that obscures the point, but preserve what an interviewee actually meant. Check levels, pauses and names. Publish a reviewed transcript alongside the audio. W3C explains why text alternatives matter; automatic transcription is a draft, not a fact-checked publication.
Practice pathway
- Episode questionRecord the decision and keep a version you can review.
- Spoken scriptRecord the decision and keep a version you can review.
- Sound testRecord the decision and keep a version you can review.
- Edit and transcriptRecord the decision and keep a version you can review.
- Cover and distributionRecord the decision and keep a version you can review.
- Listener reviewRecord the decision and keep a version you can review.
Try it
Make a three-minute pilot with one question, one speaker, two verified sources and a clear ending. Deliver the audio, final script and corrected transcript. Ask a fresh listener where they lost the thread.
Project deliverable
A reviewed piece, its source sheet and a note explaining what changed.
Extended practice: from a source record to a reviewed piece
This is a 2026 adaptation of topics documented in the former catalog. It is not a reproduced course, assessment by former faculty, enrollment offer or promise of a placement. Complete the deliverables at your own pace and review the result before publishing it.
- Episode promiseDefine the listener, question and outcome in three lines. State what the available sources cannot resolve. Test the title with someone who has not read your script.
- Spoken scriptDraft an opening, three sections and a close. Use the text timer as an estimate, then time a real read. Note the difference: breaths and changes of speaker also take time.
- Two recording testsRecord thirty seconds in two positions using the same device. Compare noise, distance and intelligibility at a comfortable volume. Keep the tests to justify your choice without treating them as technical certification.
- Interview and editingPrepare an open question and an evidence-based follow-up. Edit an answer without changing its meaning and keep the original clip. Identify any reconstruction or synthetic voice rather than presenting it as a real interview.
- Distribution and contextPrepare a title, description, credits, transcript and music provenance. Make sponsorship recognisable if present. Check the chosen platform’s requirements instead of promising income or reach.
- Review listeningDeliver a five-minute episode and its rundown. Ask a listener to summarise the main idea. Correct an unclear transition and document why it changed.
Deliverable and review criteria
Keep the first version, the reviewed piece, its source sheet and a change log. Check names, attribution, rights and a text alternative. Completing a checklist does not verify every claim in the piece.
Download a working template
A pathway, worked example and six checks in an editable file. No account.
Download TXT ↓Review before publishing
Never present a synthetic voice or a rearranged answer as an unedited spontaneous statement.
Frequently asked questions
Can I enroll through this page?
The 2024–25 catalog informs today’s self-directed guides. No enrollment or accredited qualifications are offered.
What do I need to practise?
Start with this guide’s exercise and template. Use a document for sources and, for audio or video, a device that can record a test you can review. Choose material you have permission to use.
How do I review the result?
Make a three-minute pilot with one question, one speaker, two verified sources and a clear ending. Deliver the audio, final script and corrected transcript. Ask a fresh listener where they lost the thread.
Sources and provenance
- W3C WAI: Making Audio and Video Media Accessible
- Reuters Journalistic Standards
- OBS · recording and audio setup
- Catalog · 2025
The example and pathway are new work; historical sources document the former topic, not a current offer.
A remote interview with a usable fallback
A remote interview needs an editorial plan and a route for recovering the conversation if the connection fails. Decide who monitors the recording, how the participants can signal a problem and which alternative channel is acceptable. A calendar invitation does not settle those questions.
Before the guest arrives
Record the interview's purpose, expected duration and publication format. Confirm what will be recorded and how it may be used. Check the selected platform's actual recording settings, permissions and destination file; do not assume a visible call means a usable recording exists. An alternative phone number belongs in a private production note, not in the published transcript.
Use headphones where appropriate and make a short test on the real equipment. Listen back for intelligibility, room noise and both participants' voices. Ask the guest to say a complete sentence rather than judging the connection from a greeting. The steps are a production exercise; we have not tested your microphone or connection.
A fictional twenty-minute interview
An interviewer plans a twenty-minute conversation about a community radio project. The producer checks the primary recording before the first substantive question. The agreed signal for a problem is a spoken “pause”. If the guest's audio disappears, the interviewer stops the question, notes the interruption and confirms whether a permitted alternative channel is available.
The producer does not fill the missing answer from memory. After reconnecting, the interviewer briefly states the question again and asks the guest to repeat the answer. The production log marks the repeated passage. If a usable recording cannot be made, the team arranges another conversation or changes the format with the guest's agreement; it does not publish an invented quotation.
A rehearsal with clear acceptance checks
- Make a short recording and confirm that the file opens.
- Listen to both voices using the intended review setup.
- Rehearse the pause signal and a deliberate reconnect.
- Check where the log records the interruption and repeated answer.
- Confirm the fallback is permitted and has its own recording check.
- Keep the original recording, transcript corrections and rights note together.
The worksheet separates equipment, consent, recovery and editorial evidence. A fallback is not permission to record secretly, and a transcript is not evidence of words absent from the recording. BBC Academy gives wider interview-preparation context; the fictional scenario and checklist are our independent training exercise. No student assessment, accreditation or admission is implied.
Steps: Primary recording and consent → Pause, log and confirm the fallback → Repeat the answer; verify the file
| Field | Completed example |
|---|---|
| Purpose and duration | Fictional community-radio interview; 20 minutes |
| Recording agreement | Confirm before recording |
| Equipment / test file | Short file must open and contain both voices |
| Pause signal | Say pause |
| Permitted fallback | Alternative channel only if agreed and tested |
| Log and transcript correction | Mark interruption; repeat, do not invent the answer |
Blank worksheet and worked example ↓
Primary context: BBC Academy: preparing for an interview
A five-minute episode with room to breathe
A running order allocates time; it does not prove a reading lasts exactly that long. This original exercise proposes a piece about checking a source. Record and time your own version before publishing; no real voice was measured for this example.
- Write one question that five minutes can answer. Separate speech, permitted quotations, silence and licensed music; do not allocate the same second to two different blocks.
- Treat the running order below as an editorial limit. Draft the opening and conclusion first, then include only the necessary facts and the source link in the transcript.
- Rehearse with the real equipment. Time each block from the recording rather than a words-per-minute estimate. If it overruns, cut a secondary idea before speeding up the entire reading.
- Correct the transcript against the final audio, identifying speakers and meaningful sounds. Deliver audio, a script with actual timings, the transcript and a rights record. Mark every claim that remains unverified.
| Field | Example or check |
|---|---|
| 0:00–0:20 · opening | “A screenshot can look like a source. Today we will check what is missing.” Question: can another person locate the original document? |
| 0:20–1:20 · situation | Describe a fictional screenshot without a title or date. Do not invent an institution or present it as a real incident. |
| 1:20–3:20 · procedure | Find the publisher; open the original; compare date, title and passage; record what remains unchecked. Leave a short pause between actions. |
| 3:20–4:30 · example and limit | Contrast “I saw it in an image” with “I opened this document and located this passage”. A located source can still have limitations. |
| 4:30–5:00 · close | “Keep the link and passage, not just the screenshot.” Repeat the question and offer the accessible text with sources. |
| Timing check | 20 + 60 + 120 + 70 + 30 = 300 seconds. Include pauses inside those blocks. Actual durations: __ / __ / __ / __ / __. |
Do not add filler music to hide an incomplete script. Five minutes is a production proposal, not an accredited duration or an enrolment offer from the historical catalogue.
Sources: W3C WAI — audio transcripts
Mark music, voice and effects without counting the same seconds twice
A mixing script needs to say what enters, what exits and what happens at the same time. Track durations cannot simply be added: music under speech shares its seconds. This is a fictional twenty-second production plan, not a timed recording.
- Give each resource an identifier and check its rights before use. For M01 (music) and E01 (effect), record the file, creator, source, licence or permission, permitted editing, credit and evidence of permission. A filename does not establish rights.
- Mark each track’s entry and exit alongside the spoken text. State whether a sound is alone or under speech. Do not prescribe a universal level: check that the words remain intelligible in the actual mix.
- Record the reading and measure its timing. Adjust the script if it does not fit; do not present the planned duration as measured speech.
- Check the transcript against the final audio, including sounds needed to understand it. If a resource changes or its permission remains unconfirmed, remove it or replace it with silence and listen again.
The complete exercise script
- 00:00–00:02: M01 alone; no speech yet.
- 00:02–00:06: M01 under voice: “A screenshot is not the document.”
- 00:06–00:12: M01 exits; voice: “Find the title, date and passage in the original.”
- 00:12–00:13: E01, a planned short signal; no voice.
- 00:13–00:14: intentional silence.
- 00:14–00:18: voice: “Keep the link and what remains unverified.”
- 00:18–00:20: M01 alone; exit at the end.
Check: voice 10 + 4 = 14 s; music 6 + 2 = 8 s; effect 1 s. Voice and music share 4 s: 14 + 8 + 1 − 4 + 1 of silence = 20 s. M01/E01 are placeholders in the plan, not obtained files or music with rights already confirmed.
Planned transcript: [Intro music]. Voice: “A screenshot is not the document. Find the title, date and passage in the original.” [Short signal]. [Pause]. Voice: “Keep the link and what remains unverified.” [Outro music]. After recording, correct these descriptions to match the real sound; a cue list does not replace that review.
Mixing sheet: script version ___; resource/permission ___; entry/exit ___; alone or under speech ___; measured reading ___; change needed ___; checked transcript ___; resource still to remove ___. WAI explains which speech and sound information belongs in a text alternative; it does not certify this exercise.
