Turn every sound into searchable text.
Free dictation built into iPhone and Mac, per-hour professional engines, and LLMs that follow context — pick the right one for what you're transcribing.
The engine family
Three kinds, three jobs
No tech needed — meet the three roles first, then picking by task is easy.
The clerk built into your phone
Local fast engines
Dictation built into iPhone and Mac — audio never leaves the device, works offline.
iOS · macOSThe specialist built for speech
Cloud speech engines
Cloud engines built for speech — long recordings, multi-speaker audio, domain terms.
Cloud · all platformsThe scholar who follows context
LLM transcription
Model-based transcription that works best for context-heavy cleanup, long audio, and multilingual mixed speech; dialect coverage depends on the specific model.
Cloud · all platformsSide by side
What are you transcribing?
Pick a more specific task and the table flags the engines that fit Ariya best. Scores derive only from real capabilities and price — no ranking, no single pick.
Data updated July 2026
Engine
Languages
Cost
Free tier
LocalOffline · always free · audio stays on device
Apple AccurateiOS 16 / macOS 13; runs after recording or file import; wider OS compatibility3
5
Free forever
Apple IntelligentNeeds OS 26; punctuation / word timeline / long context4
5
Free forever
Cloud STTStreaming · long audio · many languages
Azure SpeechWidest locale set (155), streaming + batch; the only Ariya engine with live translation5
3
5 hrs/mo
Aliyun SpeechStrong Chinese dialect coverage, low mainland latency; Ariya currently exposes zh / en / ja1
3
2 hrs/day · 3-mo
LLM transcriptionSemantics · long audio · multilingual mixed speech
Qwen SpeechFriendly to multilingual mixed speech, batch word-level timeline, ≤ 2h, very cheap4
5
10 hrs (90d)
Volcengine ASRStrong for Chinese, dialects, and zh/en mixed speech, low-latency streaming, files ≤ 5h4
5
Trial credit
Scores derive only from facts, mapped 0–5 to bar length. Language coverage is based on the languages and dialects Ariya currently exposes, then checked against each provider's public docs (Azure 155 · Volcengine 25 · Qwen / Apple Intelligent 21 · Mistral 13 · Apple Accurate 12 · Deepgram 10 · AssemblyAI 6 · Aliyun / AmiVoice 3). Cost = mapped from each engine's per-audio-hour price at its cheapest usable tier — cheaper scores higher (free = 5). The SDK has no structured accuracy / diarization / word-timestamp fields, so those are not scored axes. Real-time streaming is broadly available here, so it doesn't separate engines. Free tiers and prices are sample estimates; verify on each provider's site.
Privacy · Pricing
The key is in your hands.
On-device dictation never leaves the device; cloud engines use your own API key — recordings never route through us, and you pay the provider for what you use.
Local enginesApple Accurate · Apple Intelligent, no signup
Free foreverCloud enginesBring your API key; each has a free tier or new-user credit
See table aboveTransparent billingPay the provider for what you use; key stays on your device
No markupFree tiers per each provider's site.
Common questions
Still unsure?
Is an API key hard to get?
Think of it as a membership card: sign up, copy, paste. Every engine has a setup entry; most take under ten minutes, and the slow part is just signup verification.
How far does free get me?
On-device engines are free forever and unlimited (Apple devices only). Every cloud engine has a free tier or new-user credit — see the “Free tier” column above; beyond that you pay the provider for what you use with your own key, no markup.
Are my recordings and keys safe?
On-device recordings never leave the device. Cloud engines connect from your device straight to the provider; your key lives in the system secure store and can be replaced or deleted anytime.