Which API should I use?

Velma-2 offers several endpoints with overlapping capabilities. This guide helps you pick the right one based on your latency needs, language requirements, audio format constraints, and required features.

​

Quick reference

​

Speech-to-text (STT)

​

Synthetic voice detection (SVD)

​

PII/PHI redaction

PII/PHI tagging (STT APIs) — wraps sensitive spans in tags within the transcript text; original content is preserved. PII/PHI audio redaction (Redaction APIs) — replaces each detected PII/PHI span with an entity-type tag (e.g. [FIRSTNAME], [SSN], [PHI]) in the transcript text and silences the corresponding audio ranges in the returned MP3.

​

Language detection

Language detection is a pure classification endpoint — it returns no transcript, diarization, or enrichment data.

​

Decision tree

​

Do you need transcription?

Yes → continue below. No, you need deepfake detection only → use the Synthetic Voice Detection batch endpoint for pre-recorded files, or the streaming variant for live audio. See the API Reference tab. No, you need language identification → use the Language Detection batch endpoint. It returns the detected language as an ISO 639-1 code and display name with a confidence score. No transcription, diarization, or enrichment data is returned. See Language detection above.

​

Do you need PII/PHI removed from the audio itself — not just tagged in the transcript?

Yes → use the PII/PHI Redaction Batch API for complete files, or the PII/PHI Redaction Streaming API for live audio. Both return a redacted MP3 with sensitive audio ranges silenced, alongside a transcript where each PII/PHI span is replaced with an entity-type tag. No, you only need the transcript tagged → continue below, and enable pii_phi_tagging=true on whichever transcription endpoint you choose.

​

Do you need results in real time, while audio is still being captured?

Yes → choose based on your requirements:

• Need enrichments or multilingual support → use the STT Streaming API. Delivers per-utterance results with speaker labels, emotion, accent, and PII/PHI tagging.
• English only, pure transcription, lowest latency → use the STT Streaming v2 API. Emits a rolling partial transcript every ~1.5 seconds and a single final transcript at end-of-stream — no enrichments.

No, you have a complete file → continue below.

​

Is the audio English-only, and do you need maximum throughput without speaker or emotion metadata?

Yes → use the STT Batch English VFast API.

• Fastest option for English audio.
• No speaker diarization, emotion, accent, PII/PHI tagging, or utterance-level data.
• Returns a single text string and duration_ms.
• Processing timeout: 60 seconds per request.

No → use the STT Batch API.

​

STT Batch vs STT English VFast — detailed tradeoffs

​

Do you need deepfake detection alongside transcription?

The STT Batch API supports a deepfake_signal parameter that adds a per-utterance deepfake_score to transcription output. This is convenient when you already need transcription and want a synthetic voice signal without a second API call. Use the dedicated Synthetic Voice Detection APIs instead when:

• Transcription is not needed — you only want to know if audio is synthetic.
• You need frame-level results across the full audio (not just utterance-level scores).
• You need explicit no-content verdicts for silent regions.
• You need streaming deepfake verdicts in real time.

​

Common scenarios

Call center QA on recorded English calls, high volume → STT Batch English VFast. Use parallel requests with a semaphore to respect concurrent limits. Meeting transcription with speaker attribution and emotion → STT Batch with speaker_diarization=true and emotion_signal=true. Live interview transcription with real-time captions, multilingual or with speaker labels → STT Streaming. Stream audio over WebSocket and display utterances as they arrive. Live English captions or voice assistant input with minimal latency → STT Streaming v2. Partial transcripts update every ~1.5 seconds — replace your displayed text with each new partial, then finalize on the utterance message. Detect AI-generated voice in a submitted audio clip → Synthetic Voice Detection batch endpoint. Real-time anti-spoofing check during a voice authentication flow → Synthetic Voice Detection streaming endpoint. Stream audio and act on frame verdicts as they arrive. Transcript of a support call with PII/PHI spans tagged for downstream review → STT Batch with pii_phi_tagging=true. Language is auto-detected per utterance. The transcript content is preserved; sensitive spans are wrapped in tags. Compliance recording that must be shareable with PII/PHI silenced → PII/PHI Redaction Batch API. Returns a redacted MP3 with sensitive audio ranges silenced and a transcript where each PII/PHI span is replaced with an entity-type tag. Route audio to the right transcription pipeline based on spoken language → Language Detection batch endpoint. Send the clip, read predicted_language_code from the response, and route accordingly. Use the confidence field to catch low-certainty results and handle them separately.

​

Related

• Authentication and rate limits
• Audio formats
• STT enrichment features