OpenAI tts-1-hd at $30 per million characters vs Cartesia Sonic

On the listed prices, OpenAI's tts-1-hd costs $30 per million characters and tts-1 costs $15, while Cartesia's Sonic costs between $37.38 and $65 per million characters depending on your plan. So OpenAI is cheaper per character at list. Through Sume's TTS Router, Sonic-3.6 works out to $47.50 per million characters before the platform fee. Price is only part of the choice: the models are different products.

OpenAI and Cartesia list prices

Cartesia bills in credits, and its pricing page says one credit equals one character and a minute of audio takes 750 to 800 credits. That makes its plan prices convertible to a per-million-character rate, shown below. The per-million figures for Cartesia plans are arithmetic on the page's plan prices and credit allowances.

What a minute of speech costs

Using Cartesia's own ratio of 750 to 800 characters per audio minute, and assuming the same ratio for OpenAI text (which OpenAI does not state), a minute of speech is about 750 x $0.00003 = $0.0225 to 800 x $0.00003 = $0.024 on tts-1-hd, and half that on tts-1. At Cartesia's $38 per million, a minute is $0.0285 to $0.0304. Either way, voiceover is cents per minute and the rate is rarely the budget driver; retakes are.

What Sume charges for Sonic

Sume's TTS Router takes model: sonic-3.6 on POST /v1/tts-router/generate. Billing is per character: Cartesia's list price per credit is $38 per million, times Sume's 1.25 policy multiplier, which is $47.50 per million or $0.0475 per 1,000 characters, and the job total is rounded up to the next cent. A 1,000-character script is $0.0475 before rounding and the platform fee, so it bills as $0.05. The router is Cartesia Sonic only. It does not include OpenAI or ElevenLabs voices, and the first version has no streaming TTS.

• Script of 5,000 characters: $0.2375 at Sume's rate, billed as $0.24 after rounding up to the cent.
• Limits: 1 to 20,000 characters per transcript, one voice per job.
• Voice clones stay free to create under Sume's pricing policy.

How to choose

If you only need cheap, serviceable narration and already use OpenAI, tts-1 or tts-1-hd are the cheapest rows here. Choose Sonic when you want its voices or cloning, and Sume's route when you want the speech in the same job, billing and artifact flow as video and lip-sync. No quality benchmark was checked for either, so listen to both on your script before deciding.