Cloudflare Workers AI TTS: Aura-1, Aura-2, MeloTTS vs Sume TTS

On Cloudflare Workers AI, Aura-1 text-to-speech costs $0.015 per 1,000 characters and Aura-2 costs $0.030, against $0.0475 on Sume TTS 1.0, so a million characters is $15, $30 or $47.50. Cloudflare also lists MeloTTS at $0.0002 per minute of audio, which cannot be converted to a per-character figure without assuming a speaking rate. All Cloudflare numbers are from its pricing page read on 2026-10-05.

The rates

The Cloudflare page prices TTS two ways. Aura-1 and Aura-2 are per 1,000 characters, and MeloTTS is per minute of generated audio. Sume TTS 1.0 is $0.0475 per 1,000 characters, which is the provider list of $0.038 times 1.25, from the Sume pricing page. The table prices the three character-based rates at three monthly volumes.

Why MeloTTS is left out of the table

A per-minute rate depends on how fast the voice speaks, and this post has no measured speaking rate for either service, so converting $0.0002 per minute to characters would be a guess. If you use MeloTTS, generate a few of your own scripts, note the minutes of audio each produces, and multiply by the rate. Its price per minute is very small, and the better question is whether the voice and language support fit, since that is where the choice is made.

What the gap is worth

Against Aura-2 the gap is $17.50 per million characters. At 100,000 characters a month it is $1.75 and at 10 million it is $175. Against Aura-1 the gap is $32.50 per million, or $3.25 per 100,000 characters. These are not the same voices, and neither the price nor this table says which one sounds better for your content.

Sume enforces a limit of 20,000 characters per TTS request and bills $0.95 for a full one. Check the request limits for any Cloudflare model you are considering as well, since they affect how you split long scripts, and how many calls a large job makes. A pipeline that makes ten times as many calls has ten times as many chances to fail halfway, which is a cost the per-character price does not show.

• Aura-1: 68 percent below Sume per character.
• Aura-2: 37 percent below Sume per character.
• MeloTTS: priced per minute, no like-for-like figure.

A worked example

Take an audiobook sample of 8,000 characters, narrated as a single script. Aura-1 would cost $0.12, Aura-2 $0.24 and Sume $0.38. A catalog of 500 such samples is $60, $120 and $190 respectively.

Those totals are small next to the cost of the video or the book they belong to, and small next to the hours it takes to review the audio. If you add a music bed on Sume at $0.125 and a caption job at $0.20, each sample's full audio-and-caption cost is about $0.705 at Sume's rate. The pattern repeats: speech is the cheapest line in most production budgets, so the per-character gap matters only at very high volume.

Choosing

If you already run on Cloudflare Workers and want speech inside the same account, the Aura rates are lower than Sume's, and the integration is closer to your code, with no extra vendor to onboard. If your pipeline also generates video, images, captions and music, one Sume key and one balance across all of it can be worth the extra $17.50 per million characters. Do a blind listen with a 500-character sample before you decide on price, and listen on the speakers your audience will use.

Read Cloudflare's pricing page for the live rates, and the Sume pricing page for ours.