Transcriwise is not locked to a single provider. Select manually or let the system pick the engine best suited to the context — volume, language, domain, and privacy.
Multi-ASR ecosystem
Whisper Large V3 Turbo
File processing
AssemblyAI
Files and Live
ElevenLabs Scribe
Files and Live
Soniox
Files and Live
Deepgram Nova-3
Files and Live
Gladia
Files and Live
Voxtral (Mistral)
File processing
iFlytek
Files and Live
Alibaba Cloud
File processing
Guided selection
The ideal provider changes with the recording and expected result. Use filters as a starting point, not a quality guarantee.
Whisper model optimized for speed
High-performance Whisper model with an integrated diarization pipeline. Ideal for long audio with technical vocabulary and results in minutes.
Best for:
Court hearings, long lectures, extended interviews
Speakers: Unlimited
High-accuracy universal model
Cloud API with a high-accuracy universal model. Stands out for speaker identification by name or role and smart result caching.
Best for:
Corporate meetings, testimony with multiple participants
Speakers: Unlimited
Advanced sound event detection
Cloud API with Scribe v2 that detects non-verbal events (laughter, music, applause) and supports up to 32 speakers. Ideal for subtitle generation.
Best for:
Subtitle generation, multimedia content, podcasts
Speakers: 32
Simultaneous transcription and translation
Async API with automatic language identification and simultaneous translation into the target language. Domain context for specialized vocabulary.
Best for:
Multilingual content, translated subtitles, international meetings
Speakers: Unlimited
Domain-specialized models
Cloud API with domain-specialized models (medical, legal, financial, engineering). Automatic smart formatting adds punctuation, paragraph breaks, and speaker turns.
Best for:
Medical reports, financial documents, formal minutes
Speakers: Unlimited
Code-switching and 100+ languages
API with native code-switching (automatic language switching), diarization, and 100+ languages. All features included at no extra cost.
Best for:
Multilingual meetings, language-switching content, international events
Speakers: Unlimited
Native diarization and context biasing
Mistral's file-transcription model with native diarization, per-word timestamps, and context biasing with up to 100 terms.
Best for:
High-fidelity transcriptions, difficult audio, technical vocabulary
Speakers: Unlimited
Specialist in Mandarin and Cantonese
Engine focused on Mandarin and Cantonese. Supports batch and live transcription with Chinese→English translation in compatible flows.
Best for:
Chinese content, conferences in China, zh→en simultaneous translation
Speakers: Unlimited
Qwen3-ASR, Paraformer, and SenseVoice
Alibaba's model family with Qwen3-ASR for multilingual content, Paraformer optimized for Chinese, and self-hosted SenseVoice.
Best for:
Low-cost Chinese, multilingual Asia, emotion detection
Speakers: Unlimited
Feature comparison
A quick view of each engine and model's declared capabilities. Availability also depends on mode and plan.
| Feature | Whisper | AssemblyAI | ElevenLabs | Soniox | Deepgram | Gladia | Voxtral | iFlytek | Alibaba |
|---|---|---|---|---|---|---|---|---|---|
| Speaker diarization | |||||||||
| Per-word timestamps | |||||||||
| Audio events | |||||||||
| Built-in translation | |||||||||
| Domain models | |||||||||
| Custom vocabulary | |||||||||
| Speaker ID by name | |||||||||
| Live transcription | |||||||||
| Available models | Turbo (default) | Universal-3 Pro (default) | Scribe v2 (default) | STT Async v4 (default) | Nova-3 (default) | Pre-recorded v2 (default) | Mini Transcribe V2 (default) | Advanced LLM (default) | Qwen3-ASR (default) |
Formatted flows may apply dictionaries, normalization, and punctuation according to mode. Raw and Live follow their own rules.
Configurable post-processing
Post-processing depends on job mode and options; it does not change the raw transcript or replace human review.
50+ automatic corrections of common errors in legal terminology
STF, STJ, TRF, TRT, OAB, LGPD, CLT, and 30+ recognized acronyms
Automatic punctuation around statute articles, paragraphs, and legal citations
Heuristics flag recurring ASR phrases for review (e.g., "thanks for watching")
Automatic routing weighs language, mode, and availability — or pick a compatible provider yourself. Usage depends on plan, duration, provider, and optional features.
