Free text-to-speech API allowances compared, October 2026

Three vendors publish a free text-to-speech allowance on their pricing pages: Azure at 0.5 million characters per month, Speechmatics with its first 1 million characters free, and Deepgram through a $200 signup credit. Sume's docs publish no free TTS allowance and no public TTS rate, so the useful comparison is limits: up to 20,000 characters per request and a 1,200-second audio ceiling.

What each page says

Rates are quoted only where the page prints one. Deepgram's Aura-2 is $0.030 per 1,000 characters, so $200 would cover about 6.7 million characters if spent only on Aura-2.

Sume TTS limits to plan around

Each request takes up to 20,000 characters, and synthesized audio longer than 1,200 seconds fails with tts_duration_exceeded. Output defaults to mp3 at 44.1 kHz and 128 kbps.

import os, requests

r = requests.post(
    "https://api.sume.com/v1/tts-1.0/generate",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}",
             "Idempotency-Key": "tts-demo-001"},
    json={"transcript": "Welcome back. Today we cover three updates.",
          "avatar_id": os.environ["SUME_AVATAR_ID"],
          "generation_config": {"speed": 0.95, "emotion": "warm"},
          "output_format": {"container": "mp3", "sample_rate": 44100,
                            "bit_rate": 128000}},
)
print(r.status_code, r.json())

How to test fairly

Use the same script on every service, include numbers and brand names, and listen for pronunciation. Sume accepts a pronunciation_dict_id for names. Free characters vanish quickly: 1 million characters is only a few hours of narration.

Worked example

Plan your evaluation around the limits.

• Take a 1,000-character script and a 15,000-character script.
• Run both on each service and listen for pronunciation and pacing.
• On Sume, remember the 20,000-character and 1,200-second limits.

Checklist before you commit

Because the free allowances are in characters, a single long script can eat most of a monthly Azure allowance: 0.5 million characters is 25 scripts of 20,000.

• Check licensing for generated audio.
• Add a pronunciation dictionary for brand names.
• Disclose synthetic voices to listeners where required.