AITOT

Calculator

Speech-to-Text & Text-to-Speech Pricing

Compare STT (per-minute) and TTS (per-1M-character) pricing across 7 providers. Plug in monthly audio minutes and characters to see the cheapest stack.

Pricing data refreshed:

The AITOT Speech Pricing calculator compares Speech-to-Text (STT) and Text-to-Speech (TTS) cost across 7 providers — OpenAI, Google Cloud, Azure AI Speech, AWS (Transcribe + Polly), Deepgram, AssemblyAI, and ElevenLabs. STT is billed per minute of audio; TTS is billed per million characters of input text — different units from every other calculator on this site.

At 10,000 minutes of batch transcription per month, AssemblyAI Universal-2 is cheapest at $25 ($0.0025/min); OpenAI GPT-4o mini Transcribe runs $30. For real-time/streaming, Deepgram Nova-3 leads at $0.0048/min ($48/month at the same volume), just ahead of ElevenLabs Scribe v2 Realtime at $0.0065/min. On the TTS side, 5M characters/month costs $20 on Google Cloud Standard/WaveNet or AWS Polly Standard voices, versus $800 on Google Cloud Studio voices — a 40× spread for the same text.

Streaming/real-time STT costs meaningfully more than batch on the same provider (Azure: 5.5×, AWS: identical, Deepgram: ~1.8×) because it needs always-on low-latency infrastructure. OpenAI is the one outlier on TTS: it bills per token, not per character, so its $/1M-characters figure in this calculator is a derived approximation, not a native rate — flagged wherever it appears.

Speech-to-Text pricing

Cheapest · monthly

AssemblyAI · Universal-2 (pre-recorded)

$0.0025 $ / minute

$25/month

ProviderModel$ / minuteTotal / mo
AssemblyAIUniversal-2 (pre-recorded)$0.0025$25
OpenAIGPT-4o mini Transcribe$0.003$30
Azure AI SpeechStandard (batch)$0.003$30
AssemblyAIUniversal-3.5 Pro (pre-recorded)$0.0035$35
ElevenLabsScribe v1/v2 (batch)$0.0037$37
OpenAIGPT-4o Transcribe$0.006$60
DeepgramNova-3 (pre-recorded)$0.0077$77
Google CloudSpeech-to-Text v2 (Standard/Chirp)

Volume discount to $0.004/min at 2M+ min/month

$0.016$160
AWS TranscribeStandard (batch)$0.024$240

Text-to-Speech pricing

Cheapest · monthly

Google Cloud · Standard / WaveNet

$4.00 $ / 1M characters

$20/month

ProviderModel$ / 1M charactersTotal / mo
Google CloudStandard / WaveNet$4.00$20
AWS PollyStandard$4.00$20
OpenAIGPT-4o mini TTS

Billed per token ($0.60/M text-input + $12/M audio-output), not per character — figure is a derived approximation from ~1,000 input chars ≈ 1 min audio ≈ $0.015

$15.00$75
DeepgramAura-1$15.00$75
Google CloudNeural2$16.00$80
Azure AI SpeechNeural$16.00$80
AWS PollyNeural$16.00$80
Azure AI SpeechNeural HD

Dropped from $30/M in March 2026

$22.00$110
Google CloudChirp 3: HD$30.00$150
AWS PollyGenerative$30.00$150
DeepgramAura-2$30.00$150
ElevenLabsFlash / Turbo$50.00$250
AWS PollyLong-Form$100.00$500
ElevenLabsMultilingual v2/v3$100.00$500
Google CloudStudio$160.00$800

Streaming/real-time STT usually costs more than batch on the same provider. OpenAI TTS bills per token, not per character — its $/1M-characters figure is a derived approximation, not a native rate.

What this calculator does

7 providers, 2 pricing units

STT in $/minute, TTS in $/1M characters — the units every real speech bill actually uses.

Batch vs streaming toggle

STT real-time/streaming pricing differs from batch/pre-recorded by up to 5.5× on the same provider.

Independent STT + TTS tables

Most apps use only one direction (transcribe calls, or narrate content) — compare each separately.

Monthly cost at your volume

Enter your real minutes/characters per month, not a generic per-unit rate.

Cheapest-first sorting

Both tables sort by total monthly cost so the winner at your scale is always row one.

Token-billed TTS flagged

OpenAI GPT-4o mini TTS bills per token — its per-character figure is clearly marked as an approximation.

Quick comparison

Monthly cost at 10,000 STT minutes (batch) + 5M TTS characters

ComponentCheapest providerMonthly
STT batch (10k min)AssemblyAI Universal-2$25
STT batch (10k min)OpenAI GPT-4o mini Transcribe$30
STT streaming (10k min)Deepgram Nova-3 (streaming)$48
TTS (5M chars)Google Cloud / AWS Polly Standard$20
TTS (5M chars)Deepgram Aura-1$75
TTS (5M chars)Google Cloud Studio$800

Standard/base voices are 8–40× cheaper than premium (Studio, Long-Form) tiers for the same character count.

How to use this calculator

Compare Speech-to-Text and Text-to-Speech monthly cost across 7 providers.

  1. 1

    Enter STT minutes/month

    Total audio minutes you transcribe monthly — calls, meetings, voicemail, video captions.

  2. 2

    Pick batch or streaming

    Batch (pre-recorded) is cheaper. Streaming (real-time) costs more but is required for live captions/voice agents.

  3. 3

    Enter TTS characters/month

    Total input text characters you synthesize to audio monthly.

  4. 4

    Compare and pick

    Both tables sort cheapest-first. Standard voices win on cost; premium (Studio, Long-Form) voices win on naturalness.

Why use this calculator

  • 7 providers, STT + TTS
  • Batch vs streaming modeling
  • Real per-minute + per-character units
  • Cheapest-first sorting
  • Token-billed TTS flagged, not hidden
  • No login required

Frequently Asked Questions

What is the cheapest Speech-to-Text API in 2026?+
For batch/pre-recorded transcription: AssemblyAI Universal-2 at $0.0025/minute ($0.15/hour), or ElevenLabs Scribe at $0.0037/minute. For real-time/streaming: Deepgram Nova-3 streaming at $0.0048/minute is currently the cheapest, just ahead of ElevenLabs Scribe v2 Realtime at $0.0065/minute.
What is the cheapest Text-to-Speech API in 2026?+
Google Cloud Standard and AWS Polly Standard voices both price at $4 per 1M characters — the cheapest tier across all 7 providers compared. Premium naturalness tiers (Google Studio at $160/M, AWS Long-Form at $100/M) cost 25–40× more for the same character count.
Why does OpenAI TTS pricing look different from other providers?+
OpenAI's gpt-4o-mini-tts bills per token ($0.60/M text-input + $12/M audio-output tokens), not per character like every other TTS provider. The calculator converts this to an approximate $/1M-characters figure (~$15) using the community-documented ~1,000 characters ≈ 1 minute of audio ratio — it is flagged as a derived estimate, not a native rate.
Is streaming Speech-to-Text more expensive than batch?+
Usually, but not universally. Azure charges 5.5× more for real-time ($1/hour) than batch ($0.18/hour). AWS Transcribe prices batch and streaming identically at $0.024/minute. Deepgram is the counterexample — its Nova-3 streaming tier ($0.0048/min) is actually cheaper than its pre-recorded tier ($0.0077/min), reversing the pattern of its previous Nova-2 generation.
How much does it cost to transcribe 1,000 hours of audio per month?+
1,000 hours = 60,000 minutes. At AssemblyAI Universal-2 ($0.0025/min): $150/month. At OpenAI GPT-4o mini Transcribe ($0.003/min): $180/month. At AWS Transcribe tier 1 ($0.024/min): $1,440/month — but AWS drops to $0.015/min above 250K minutes/month, dropping the effective cost significantly at that scale.
Should I pick a Standard or Neural/premium TTS voice?+
Standard voices (Google/AWS at $4/M characters) are fine for internal tools, accessibility read-aloud, and draft content. Neural voices ($16/M) sound noticeably more natural and are the default choice for customer-facing audio. Studio/Long-Form voices ($100-160/M) are worth it only for flagship narration (audiobooks, ads) where voice quality is the product.