Transcribes spoken audio into text.
Product
Languages
Price
~50 languages
$0.006/min (~$0.36/hr)
Legacy Whisper and GPT-4o Transcribe are both priced the same; diarization (speaker labels) included at no extra cost.
~50 languages
$0.003/min (~$0.18/hr)
Cheaper, lower-accuracy tier of the same transcription API.
125+ languages and variants
$0.016/min (~$0.96/hr)
Standard recognition, first 500K minutes/month; volume pricing steps down to $0.004/min above 2M minutes/month. A cheaper batch tier ($0.003/min, 24-hour turnaround) also exists.
99 languages
$0.22/hr
Realtime variant costs more, $0.39/hr. Entity detection (+$0.07/hr) and keyterm prompting (+$0.05/hr) are optional add-ons.
45+ languages
$0.0043–$0.0058/min (~$0.26–$0.35/hr)
Monolingual streaming $0.0048/min, multilingual streaming $0.0058/min; pre-recorded audio is slightly cheaper than streaming.
23 Indic languages + English
₹30/hour (~$0.32/hr)
Converted at ₹94.6/$1 (Sept 2026 mid-market rate) — check docs.sarvam.ai for the current ₹ price, which is the one Sarvam actually bills.
English + 8 Indic languages
Not publicly priced
An end-to-end speech LLM — understands speech directly and translates it, without a separate ASR step. Open weights on Hugging Face/GitHub; commercial API access is via Krutrim Cloud, which doesn't publish a per-minute or per-character rate.