How to Separate Interview Speakers with SpeakerSplit
SpeakerSplit processes recordings to distinguish speakers and provide speaker-related audio and transcript outputs. It suits an interviewer or editor who has a conversation recording and needs to work with its participants separately. The useful result is a reviewed editing aid, not an assumption that every overlapping voice has been isolated perfectly.
By Fırat Papatya · Published · 5 min read
Affiliate disclosure: the SpeakerSplit partner link in this article may earn Kuponaut a commission at no extra cost to you. KUPONAUT is recorded as a partner code. Prices are from Kuponaut’s pricing record of 20 September 2026 and can change. This guide is based on official documentation, not a personal product review.
Step 1: prepare the recording and keep the original
Keep an untouched copy of the recording and check that you have permission to upload it. Confirm that the file contains the conversation you intend to process, rather than a partial export or an unrelated track. If the recording contains confidential discussion, review the service's current handling terms before sending it to an external system.
SpeakerSplit's FAQ lists common audio and video formats, including MP3, WAV, M4A and MP4. Check the current uploader for the formats and size restrictions applicable to your file. Listen to the beginning and a difficult section so that you know whether the source contains background noise, interruptions or people talking together. That gives you a reference for evaluating the output later.
Step 2: choose the speaker-separation workflow
The official Speaker Split page presents Start Separating Speakers and describes uploading audio or video for speaker separation and a diarized transcript. Use that workflow for a conversation. Do not assume that every audio-separation mode solves the same problem: separating participants in an interview is different from extracting a vocal performance from music.
The FAQ describes automatic speaker detection and an option to specify a speaker count. Use the available controls according to what is actually in the recording. Count distinct participants, not microphone channels, and avoid guessing extra speakers merely because a person's voice changes volume. Check any processing allowance shown before starting; this guide does not invent a duration or cost for a job.
Step 3: inspect labels and transcript against the recording
SpeakerSplit's FAQ describes accessing completed work through History and opening the transcript, which includes speaker labels, in an editor to make changes. Review a representative passage for each participant before relying on the attribution. Correct obvious label mistakes and compare the words with the audio, especially where a name, technical expression or quiet response could have been misunderstood.
Keep uncertainty visible. If people speak over each other and you cannot confidently identify a sentence, do not assign it to someone merely to produce a tidy transcript. A readable transcript can still contain an attribution error. For quotations, return to the original recording and preserve enough context to establish what the person actually said rather than relying on an automated label alone.
Step 4: download the outputs needed for the next task
The Speaker Split page describes time-synchronised speaker tracks and transcript exports, with audio options such as WAV or MP3 and text or subtitle formats including TXT and SRT. Choose the output according to the next job: audio editing, reading or caption preparation. Do not assume that a transcript file and an isolated audio track contain the same information.
Open the downloaded files and compare their names with the participant labels you reviewed. Listen to transitions as well as clean sections. Keep the original beside the processed outputs so that missing words or unnatural edits can be checked later. Download what you need promptly: SpeakerSplit's FAQ says separated audio files are kept for 30 days, so do not treat the job history as a permanent archive of important interview material.
Tips for interviews with noise or overlapping speech
SpeakerSplit's FAQ acknowledges that source quality affects results. Reduce unnecessary noise and encourage clear turn-taking when you control the original session. For an existing difficult recording, begin with a representative passage during the trial rather than assuming an entire archive will behave like a clean demonstration. Evaluate the result for the editing task you actually need to complete.
Isolated tracks can contain artefacts or fragments from another voice. Listen before amplifying a participant or removing a section. Keep natural pauses where they matter to the meaning of an answer. Do not use separation to create a misleading exchange, and do not present an uncertain transcript as an exact account without checking the original audio.
Processing allowances and what separation cannot establish
Kuponaut’s pricing record describes paid plans by the amount of audio processed each month and a trial for a short recording; SpeakerSplit's FAQ currently describes monthly credit plans, with credits charged per file according to its duration. It does not specify a current paid amount or universal trial duration. Check those details in your account before uploading a long session. A successful upload does not establish that the available allowance covers all later processing or revisions.
The official pages are not consistent enough to support a single language-count claim here, so this guide gives none. Check the current supported-language information for your recording. More importantly, a speaker label is not identity verification. Use the tool to assist editing and review, not to prove who a person is, settle a disputed quotation or certify the accuracy of a transcript.
Recorded prices, plan allowances and the Kuponaut offer
Prices recorded on 20 September 2026: A free trial lets you process a short recording without paying. Paid plans are priced by minutes of audio processed per month; see the pricing page for current tiers. (SpeakerSplit's FAQ currently describes these plans as monthly credits charged by file duration.)
The brand’s partner programme lists KUPONAUT as 10% off every payment. The code comes from the brand’s partner programme and Kuponaut has not tested it at checkout, so check the eligible plan and final total before you pay. The recorded expiry is 20 September 2027.
Kuponaut’s record does not contain a paid price or a fixed trial-minute count, so none is given here. Confirm whether a new upload or rerun consumes the current allowance before processing a larger collection.
Speaker separation or a different audio workflow?
SpeakerSplit's own documentation distinguishes the conversation-oriented workflow from separating vocals and instrumental audio. Choose according to the source and the desired output; a music separation task is not interchangeable with attributing an interview sentence to a participant. The feature names should not replace listening to the recording and deciding what the editor actually needs.
If the original recording already contains separate microphone tracks, work from those originals where possible instead of treating a mixed export as the only source. If you only need a readable account, a manually reviewed transcript may be enough without isolated audio. This guide does not rank separation services or claim that processing restores information absent from the recording.
Related code pages
- SpeakerSplit discount code — Tested at checkout November 4, 2026: 10% off every payment.
- Zeemo discount code — Tested at checkout November 4, 2026: 10% off your first payment.
- TTS Buddy discount code — Tested at checkout November 4, 2026: 20% off all payments in the first 3 months.
- Zeemo: zeemo pricing
- How to turn a long video or podcast into short clips with Klap
- How to Add Readable Captions to a Video with Zeemo
- How to Turn Study Notes into Audio with TTS Buddy