Audio Speaker Identification Annotation -- 2

₹750-1250 INR / hour

Closed

Posted 7 days ago

₹750-1250 INR / hour

I need help turning a batch of raw recordings into clean, speaker-labelled data. Every file already carries basic metadata; what is missing is precise speaker identification. Your task is to listen through each track, mark the start- and end-times whenever a new voice begins, and assign a consistent speaker ID throughout the set. Please deliver: • A time-stamped annotation file for each recording (CSV or JSON) that lists segment start, segment end, and assigned speaker label. • The annotated audio back in its original folder structure so it stays aligned with our internal pipeline. Accuracy matters more than speed, so feel free to use diarization tools such as [login to view URL], SpeechBrain, or your own proprietary workflow—just document what you use so we can reproduce results later. When you reply, point me to past work that shows you have handled speaker identification before; a short paragraph and one or two sample links are enough.

Project ID: 40741085

About the project

5 proposals

Remote project

Active 2 days ago

Looking to make some money?

Email address

Benefits of bidding on Freelancer

Set your budget and timeframe

Get paid for your work

Outline your proposal

It's free to sign up and bid on jobs

5 freelancers are bidding on average ₹763 INR/hour for this job

@pvsysgroupinc

★★★ TOP 1% IN FREE LANCER WORLD ★★★ ★★★ 20+ Year Experience in IBD being CMD★★★ ★★★ 200+ Country Satisfied Clientele ★★★ ★Linkedin★ ★Data Entry★ ★Business Plans★★★ ★★★ Operational Strategic planner Customer Support 24*7★★★ ★★★Excel/Word Operation★★★ ★★★Chat Support★★★ ★★★Calling Support★★★ ★★★Business Plans / Marketing Strategy ★★★ * Digital Marketing★★★ ★★★Social Media Marketing ★★★ ★★★Internet Marketing ★★★ ★★★Any type of Data Projects★★★ ★★★★★★ Regards, ★★★CMD★★★ ★★★PVSYS GROUP (INDIA)★★★ YOUR HONEST and DEDICATED WORKER ★★★IF YOU THINK THEN I CAN★★★

₹750 INR in 40 days

5.0

(5 reviews)

3.5

@bbb51

Hello, I can turn your recordings into consistent speaker-labelled CSV or JSON with precise start/end timestamps and reproducible speaker IDs. I will: • run speaker diarization for the initial segmentation; • manually review every speaker change and correct overlaps, gaps, and misidentified voices; • verify speaker-ID consistency across the full batch; • return the annotations and audio in the original folder structure; • document the workflow so your team can reproduce it. I can deliver the first annotated file within 24 hours for schema approval before processing the rest. As a practical bonus, I will include a brief quality report listing ambiguous segments and confidence issues. What is the total audio duration, and do you prefer timestamps in seconds or milliseconds?

₹800 INR in 7 days

0.0

(0 reviews)

0.0

@langruijun

Hi — you need each track turned into segment-level annotation (start, end, speaker ID) delivered as CSV/JSON, files returned in their original folder structure, and the tooling documented so your team can reproduce it. I can start today. How I would run your batch: 1. Send me one recording and I will return its annotation within 4 hours, free — you check the segment boundaries and speaker consistency before committing to the rest. 2. Per track: voice-activity segmentation against the noise floor, then speaker features (log-mel plus spectral) clustered per file, with the number of speakers estimated from an internal separation score instead of fixed. Adjacent segments of the same speaker are merged, and any spot where two voices overlap is tagged for you rather than guessed. 3. Output: one CSV (start, end, speaker) per recording, JSON if you prefer, original file names and folder structure kept; plus a short run log listing the exact steps and parameters used on your files. 4. Every file gets a second pass against the audio before I hand it over. On past work: my public portfolio is transcription and document work rather than published diarization samples, so let the free first track be the evidence instead of a link. Rate: INR 765/hour, invoiced against milestones; a 10-minute track takes roughly an hour with the second verification pass. Question: how many files and what total duration, and do any tracks contain overlapping speech, or is it mostly turn-taking?

₹765 INR in 1 day