500 hours of Hinglish Scripted Speech

Speakers read prompts from a phonetically balanced script shown one line at a time, with retakes on any flubbed or misread line.

A 45-minute booth session yields roughly 300–400 prompts. One file per utterance, cut at the prompt boundary, so alignment is exact before any forced-alignment pass runs.

Yield at this style: 85–90% of recorded time is accepted. Prompt-level retakes catch problems inside the session, so very little is discarded afterwards. This is the highest-yield style we run.

• Script built for triphone coverage rather than word coverage, so rare phone contexts appear often enough to train on
• Digit strings, dates, currency and person names blocked separately, because these are where deployed ASR actually fails
• Domain lexicon injected from your product vocabulary when you supply one
• Sentence-length distribution spread deliberately, since all-short prompts produce a model that cannot handle long utterances
• Built against Hinglish specifically: Intra-sentential switching means English words carry Indian phonology, so English acoustic models mis-transcribe them
• Code-mixing handled explicitly rather than edited out — Hinglish is the code-mixing case itself. Typical urban customer-support speech is 30-60% English tokens embedded in Hindi grammar, with switching several times per utterance.

Eight to twelve weeks. Cities are staggered rather than started together, so the protocol corrections found in the first city are applied before the last one begins recording.

Six to eight batches, fortnightly, each speaker-disjoint so you can train on early batches without contaminating a later evaluation split.

Recruit by switching behaviour, not by language proficiency. Screening recordings are used to confirm speakers switch naturally rather than performing one language.

At 500 hours quotas are enforced per dialect and reconciled fortnightly. Aggregate demographics are reported per batch so drift is visible while there is still time to correct it.

Requires literate speakers comfortable reading aloud in the target script, which is the main constraint on cohort breadth and has to be actively counterweighted.

• Intra-sentential switching means English words carry Indian phonology, so English acoustic models mis-transcribe them
• Hinglish is the code-mixing case itself. Typical urban customer-support speech is 30-60% English tokens embedded in Hindi grammar, with switching several times per utterance.
• Almost no public corpus contains genuine intra-sentential Hindi-English switching with per-token language tags. This is the highest-value gap for anyone building Indian conversational AI.

• Prompt-to-audio match verified by a native reviewer; a misread line is a rejected file, not an edited transcript
• Hyperarticulation flagged — a speaker over-enunciating because they are reading produces audio that does not match deployment
• Clipping and truncation checked at both utterance boundaries
• Per-speaker prompt coverage confirmed, so no speaker silently skips a block
• Whether English tokens are written in Latin or transliterated into Devanagari must be fixed by rule, not left to annotators
• Language-ID tagging per token is required for training but is skipped by most vendors

100% technical QA, 10% content QA stratified by city, dialect, transcriber and recording condition, with automatic escalation to 100% on any failing batch.

• Reading voice: flat prosody and unnatural stress that trains a model on speech nobody actually produces
• Prompt fatigue in the back half of long sessions, where accuracy drops and pace flattens
• Speakers who are not fluent readers, which quietly biases the cohort toward higher education bands
• Script leakage across speakers, producing a corpus that memorises sentences instead of covering sounds

Cost at this band is driven by: Field and rural capture ratio — mobile rig hours cost more than studio hours; Rare dialect quotas, where the last 10% of the cohort can cost as much as the first 40%.

• Speaker duplication across cities becomes a real risk at this cohort size and needs active de-duplication against voice and ID
• Rural capture depends on weather and travel in a way studio work does not, so mobile-rig batches carry schedule variance
• Quota drift compounds across six cities unless demographics are reconciled weekly rather than at the end
• Hinglish carries 4 recognised varieties across Delhi NCR, Mumbai, Bengaluru, so the quota matrix is wider than the headline volume suggests
• Almost no public corpus contains genuine intra-sentential Hindi-English switching with per-token language tags. This is the highest-value gap for anyone building Indian conversational AI.

• WAV audio to your naming convention, with the per-file manifest
• Verbatim Devanagari + Latin transcripts with utterance-level timestamps
• Prompt ID mapped to every utterance
• Verbatim deviation flags where the speaker departed from the script
• Per-utterance SNR and duration in the manifest
• Per-speaker metadata: age band, gender, region, dialect, education band
• Consent records mapped to speaker IDs
• QA report with pass rates, rejection reasons and agreement statistics
• Speaker-disjoint train / dev / test splits on request

• ASR acoustic model baselines
• TTS voice building where a single speaker is recorded at depth
• Pronunciation lexicon and G2P validation
• Forced-alignment and phone-boundary work

Scripted audio contains no disfluencies, no false starts and no natural turn-taking. A model trained on it alone degrades sharply on real spontaneous input, so it is a baseline layer rather than a complete training set.