Speech Data Collection in Mumbai

The largest concentration of Indian consumer-app and fintech teams, so Mumbai speech is what a lot of deployed products actually meet first.

Migration from across Maharashtra, Uttar Pradesh, Bihar and Gujarat means almost any Indian language has a resident pool, but native-region screening is essential — a Mumbai-resident Bhojpuri speaker often carries two decades of Marathi and Hindi contact.

• Marathi
• Hindi
• Hinglish
• Gujarati
• Urdu
• Indian English

Bambaiya Hindi-Marathi contact speech with constant three-way switching; the densest code-mixing environment in India.

Migrant-heavy, so speakers of almost any Indian language can be found, but native-region screening is essential.

Over-represents urban, code-mixing, higher-exposure speakers. A Mumbai-only Marathi cohort will under-represent Vidarbha and Marathwada forms almost entirely.

City choice is a data-quality decision, not a logistics decision. It determines which dialects end up in your corpus.

Treated booths in the western suburbs with parallel session capacity for multi-speaker conversation work.

Studio capacity is in the western suburbs, and cross-city travel time is the binding constraint on session throughput. Sessions are scheduled in two blocks a day rather than continuously, and monsoon months compress field work sharply.

Dense street noise, local-train ambience and small-shop interiors — genuinely the noisiest field environment in the network, and the most representative of deployed mobile use.

• 100% of delivered files pass automated technical QA for SNR, clipping, duration and silence
• 5-25% of files pass a second native-speaker content review, stratified by city, dialect and transcriber, and escalating to 100% on any batch that fails the agreed threshold
• Accepted yield runs 85-90% for scripted speech, 60-70% for spontaneous, 55-65% for conversational and 50-60% for telephony
• Default cohort quotas: 50/50 gender, with age bands at 30% (18-25), 40% (26-40) and 30% (41-60)
• 48 kHz / 24-bit capture, delivered as 16-bit PCM WAV, with studio sessions held below a -50 dBFS noise floor
• First response within one working day; a scoped, fixed quote within two to three

These are the figures a delivery is measured against, not aspirations. A batch that misses them is re-recorded at our cost rather than repaired.

• Local coordinators recruit and screen against your quota matrix
• Participants are briefed and consented on site
• Sessions run to the same template as every other city in the network
• Technical QA runs on upload, so defects are caught while the speaker can still be recalled