# 6 Fish Audio alternatives in 2026 for text to speech on a phone line

By Samuel Oram-Jones

Fish Audio prices its s2.1-pro model at $15 per million UTF-8 bytes, as of November 2026, and its text to speech API returns four formats: mp3, wav, pcm, and opus. A phone call carries 8 kHz mu-law, the 8-bit encoding North American phone networks use, so a voice agent built on Fish Audio converts its audio before a caller hears it.

Six providers return that phone format directly: Bland Speech, ElevenLabs, Cartesia, Inworld, MiniMax, and Smallest AI.

## Why do teams look for Fish Audio alternatives?

A team looks for a Fish Audio alternative when a phone line needs 8 kHz mu-law audio or a finished SOC 2 report, because Fish Audio's text to speech formats are mp3, wav, pcm, and opus, and its SOC 2 audit is still underway.

Fish Audio, from Hanabi AI Inc., is a voice AI platform with a web app, a REST API, and SDKs. Fish Audio's recommended production model is s2.1-pro, which Fish Audio documents in 83 languages.

On paperwork, Fish Audio says a SOC 2 audit "is currently underway, and the report will be available to customers under NDA once it is complete". Fish Audio also says it supports "HIPAA-aligned configurations" and can sign a BAA (a business associate agreement, the contract HIPAA requires for health data) for qualifying healthcare workloads.

A team that self-hosts the open S2 weights meets Fish Audio's Community License Agreement of 13 October 2026, which makes commercial use free only while annual revenue is under USD 500,000.

## What should I check in a Fish Audio alternative?

A team comparing Fish Audio alternatives should check which phone formats the API returns, which compliance status is already published, and how the price is counted, because the six providers answer those questions differently.

Phone formats come first because a call doesn't carry a studio file. Bland Speech, ElevenLabs, Cartesia, Inworld, MiniMax, and Smallest AI each document an 8 kHz mu-law output, so a team can request the format a carrier expects without converting audio after the fact.

Compliance is published unevenly. Bland (the company), ElevenLabs, Cartesia, and Inworld each state SOC 2 Type II status. Fish Audio's audit is underway, and Smallest AI names SOC 2 among its Enterprise features.

The price unit matters more than the headline number. Fish Audio bills per UTF-8 byte, which equals one character for plain English text and runs to two or three bytes for many accented and non-Latin letters. Cartesia bills credits at 1 credit per character against a monthly allowance.

## What are the best Fish Audio alternatives?

The best Fish Audio alternatives are Bland Speech, ElevenLabs, Cartesia, Inworld, MiniMax, and Smallest AI, with Bland Speech first because it returns 8 kHz mu-law audio, bills one standard per-character rate with no monthly plan, and ranked first of 13 AI voices in Design Arena's Audio Realism Bench results of 4 August 2026.

| Provider | Phone audio | Published price | Clone audio needed |
| --- | --- | --- | --- |
| Bland Speech | Mu-law at 8 kHz, PCM at 8 to 48 kHz | $0.04 per 1,000 characters, pay as you go | One clean sample of about 10 seconds; 30 minutes or more for a professional clone |
| ElevenLabs | ulaw_8000 and alaw_8000 | $0.08 per 1,000 characters on Eleven v4; $0.04 on Eleven v4 Turbo | 10 seconds on v4 (Starter, $6 a month) |
| Cartesia | Mu-law and A-law at 8 kHz | Pro $5 a month for 100K credits; 1 credit per character | 10 seconds (Pro); 30 minutes for a professional clone (Startup, $49 a month) |
| Inworld | Mu-law and A-law at 8 kHz | $25 per million characters on Realtime TTS-2 (On-Demand) | 3 seconds, on every plan |
| MiniMax | Mu-law at 8 kHz | $100 per million characters on speech-2.8-hd | 10 seconds to 5 minutes, plus $1.5 per clone at first use |
| Smallest AI | Mu-law, 8 kHz to 44.1 kHz | About $19.50 per million characters on Lightning v3.1 Pro | 5 to 15 seconds |

<details><summary>What these rows mean</summary>

- **Phone audio:** the telephone-ready output each provider documents.
- **Published price:** each provider's own rate, with the plan or model named.
- **Clone audio needed:** the recording length for an instant clone, and the plan that gates it where one is named.

</details>

### 1. Bland Speech

Bland Speech is Bland's own text to speech model, sold through a studio and an API that streams over HTTP chunks or a WebSocket. The WebSocket takes text as an LLM writes it, and a new context ID interrupts the turn that is playing, which is how barge-in (a caller talking over the agent) works. Bland publishes 199ms to first audio. Tags such as [laughs] and [say warmly] sit before the words they shape on Bland's current-generation voices, and an instant clone takes one clean sample of about 10 seconds. More on the [text to speech API](/text-to-speech-api).

Bland Speech suits a team that wants a hosted voice for phone calls with one rate that is the same in the studio and the API. It costs $0.04 per 1,000 characters, or $40 per million, pay as you go with no monthly plan, and a new account gets 50,000 free characters once.

<div class="cmp-proscons cmp-proscons-own"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Mu-law output at 8 kHz for phone lines, plus PCM from 8 to 48 kHz</li><li>One rate in the studio and the API, with no monthly plan</li><li>A WebSocket that takes LLM tokens and handles barge-in</li><li>An instant clone from about 10 seconds of audio</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>More than a hobby project needs, because Bland Speech is built for production calls</li></ul></div></div>

### 2. ElevenLabs

ElevenLabs sells Eleven v4 and Eleven v4 Turbo, both launched on 28 September 2026, and bills its API in US dollars per character: $0.08 per 1,000 characters on v4 and $0.04 on v4 Turbo. ElevenLabs returns ulaw_8000 and alaw_8000 among its output formats and documents 90+ languages for the v4 family. An instant clone needs 10 seconds of audio on v4 and a Starter plan at $6 a month. ElevenLabs suits a team that wants a wide language set and audio tags such as [whispering] on Eleven v4, and will manage plan tiers to get cloning.

<div class="cmp-proscons" data-rival="elevenlabs"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>90+ languages on the Eleven v4 family</li><li>Mu-law and A-law output at 8 kHz</li><li>SOC 2 Type 2, ISO 27001:2022, and PCI DSS 4.0.1 Level 1</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>The standard text to speech WebSocket takes neither Eleven v3 nor Eleven v4</li><li>Free workspaces are limited to non-commercial use</li><li>$0.08 per 1,000 characters on Eleven v4</li></ul></div></div>

### 3. Cartesia

Cartesia sells Sonic 3.6, released on 27 August 2026 in 44 languages, and says first audio arrives in under 90ms. Cartesia bills credits at 1 credit per character: Pro is $5 a month for 100K credits, Startup is $49 for 1.25M, and Scale is $299 for 8M, with overage at $65, $45, and $38 per million credits. Cartesia returns mu-law and A-law at 8 kHz. Cartesia suits a team that wants a first-audio figure under 90ms as Cartesia states it and will manage a credit allowance.

<div class="cmp-proscons" data-rival="cartesia"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Mu-law and A-law output at 8 kHz</li><li>44 languages on Sonic 3.6</li><li>Professional clone training is free</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Speech draws on a monthly credit allowance, with overages billed at renewal</li><li>A professional clone starts on a $49 plan with 2 slots</li></ul></div></div>

### 4. Inworld

Inworld sells realtime voice AI aimed at consumer apps, with Realtime TTS-2 at $25 per million characters and TTS-2 Flash at $15 on its On-Demand plan, which costs $0 a month. Inworld says first audio arrives in as low as 50 ms for TTS-2 Flash. Instant cloning starts from 3 seconds of audio on every plan, and Inworld returns mu-law and A-law at 8 kHz. Inworld suits a team building a companion or game voice that wants plain-language direction of delivery and a low per-character rate.

<div class="cmp-proscons" data-rival="inworld"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Instant cloning from 3 seconds on every plan</li><li>SOC 2 Type II, with HIPAA and GDPR compliance and a BAA on Enterprise</li><li>Mu-law and A-law output at 8 kHz</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Professional cloning is a beta that runs on TTS-2 only</li><li>Zero data retention leaves clone samples out</li><li>The REST endpoint takes up to 2,000 characters per request</li></ul></div></div>

### 5. MiniMax

MiniMax is a general AI company that sells speech next to language, video, and music models, and its current speech models are speech-2.8-hd at $100 per million characters and speech-2.8-turbo at $60. MiniMax documents 40 languages, a bidirectional WebSocket that accepts LLM output and can be cancelled mid-turn, and 19 inline interjection tags on the 2.8 models. Rapid cloning takes a clip of 10 seconds to 5 minutes and charges $1.5 per voice the first time the voice is used. MiniMax suits a team that clones many short-lived voices and accepts the per-voice fee.

<div class="cmp-proscons" data-rival="minimax"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>A bidirectional WebSocket that supports cancelling a turn</li><li>40 languages on the 2.6 and 2.8 models</li><li>Mu-law output at 8 kHz</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>$100 per million characters on speech-2.8-hd</li><li>$1.5 per cloned voice at first use</li><li>A cloned voice is deleted after 7 days without use</li></ul></div></div>

### 6. Smallest AI

Smallest AI sells a realtime voice suite for enterprise, and its text to speech product is Lightning. Lightning v3.1 Pro costs about $19.50 per million characters, pay as you go starts with $10 in free credits, and the Lightning API returns PCM, WAV, MP3, or mu-law at 8 kHz up to 44.1 kHz. Smallest AI says "sub 100ms latency", and Smallest AI's own benchmark puts first byte at about 200 ms with 40 concurrent requests. Smallest AI suits a team that wants a low per-character rate and will accept a model schedule that changes: lightning_v3.1 retires on 30 November 2026 at 18:30 UTC.

<div class="cmp-proscons" data-rival="smallest-ai"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>About $19.50 per million characters on Lightning v3.1 Pro</li><li>HTTP, SSE, and WebSocket streaming</li><li>An instant clone from 5 to 15 seconds of audio</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Professional cloning is an Enterprise feature</li><li>A HIPAA zero data retention add-on at $1,000 a month on pay as you go</li><li>lightning_v3.1 retires on 30 November 2026</li></ul></div></div>

## Which Fish Audio alternative sounds most human?

Among the alternatives that were scored, Bland Speech v3 sounded most human, ranking first of 13 AI voices at 1365 in Design Arena's [Audio Realism Bench results of 4 August 2026](https://notes.designarena.ai/audio-realism-benchmark-measuring-realism-in-audio-models/), while a human recording scored 1500.

In those results Cartesia's Sonic 3.5 ranked 6th at 1096, MiniMax's Speech-02 HD ranked 7th at 1093, Smallest AI's Lightning v3.1 Pro ranked 11th at 952, ElevenLabs' Eleven v3 ranked 12th at 938, and Inworld's TTS-1.5 Max ranked 13th at 763. Fish Audio wasn't in those results. Eleven v4 (28 September 2026), Cartesia's Sonic 3.6 (27 August 2026), and Fish Audio's Drama 3 (13 October 2026) were released after those results and aren't scored in them.

One catch applies to every number here: Design Arena scored one female and one male voice per model, each chosen as its most naturally conversational American-English voice, so the scores describe those voices and not a team's own clone.

## Which Fish Audio alternative should I pick?

Replacing Fish Audio on a phone line starts with the audio format, and all six alternatives here return 8 kHz mu-law, so the tiebreaker is how the voice sounds. On that count Bland Speech is the pick for phone calls and voice agents: Bland Speech v3 scored 1365 in the blind tests Design Arena published on 4 August 2026, 269 points above Cartesia's Sonic 3.5 and 272 above MiniMax's Speech-02 HD, the next two alternatives in those results.

Price doesn't make the case for Bland Speech. At 1 million characters of plain English text a month, published rates come to $15 on Fish Audio s2.1-pro, about $19.50 on Smallest AI Lightning v3.1 Pro, $25 on Inworld Realtime TTS-2, $40 on Bland Speech and on ElevenLabs Eleven v4 Turbo, $60 on MiniMax speech-2.8-turbo, and $80 on ElevenLabs Eleven v4, with Cartesia's Startup plan at $49 a month for 1.25M credits. Phone minutes and taxes are left out of every figure. Bland Speech costs more than Fish Audio, and the case for it rests on the realism result and phone-ready audio.

[Bland Speech vs Fish Audio](/compare/bland-speech-vs-fish-audio) goes deeper on the head-to-head, the [Fish Audio review](/blog/fish-audio-review) covers Fish Audio on its own, and [ElevenLabs vs Fish Audio](/blog/elevenlabs-vs-fish-audio) puts two of these providers side by side. The [alternatives guides](/compare/text-to-speech/alternatives) cover other providers, the [text to speech hub](/compare/text-to-speech) lists every comparison, and [Bland Speech](/speech) has a studio for a first test.

## Frequently asked questions

### What is the best Fish Audio alternative?

Bland Speech is the best Fish Audio alternative for phone calls and voice agents, because it returns 8 kHz mu-law audio, bills $0.04 per 1,000 characters with no monthly plan, and ranked first of 13 AI voices in Design Arena's results of 4 August 2026.

### Is there a cheaper Fish Audio alternative?

No, none of the six undercuts Fish Audio's $15 per million UTF-8 bytes on published rates, though Inworld's TTS-2 Flash matches it at $15 per million characters. Smallest AI's Lightning v3.1 Pro is about $19.50 per million characters and Inworld's Realtime TTS-2 is $25, and the other four cost more still.

### Which Fish Audio alternative outputs phone-ready audio?

Bland Speech, ElevenLabs, Cartesia, Inworld, MiniMax, and Smallest AI each document an 8 kHz mu-law output. Fish Audio documents mp3, wav, pcm, and opus for its text to speech API.

### Which Fish Audio alternative has the lowest published latency?

Inworld publishes the lowest figures of the six, as low as 50 ms to first audio for TTS-2 Flash and 20 ms measured server-side without network, and each company measures its own way. Bland publishes 199ms to first audio, Cartesia says under 90ms, ElevenLabs says a median of about 100ms for Eleven v4 Turbo excluding application and network latency, and Smallest AI says sub 100ms. Fish Audio says "<70 ms" without labelling it time to first audio.

### Can a Fish Audio cloned voice move to another provider?

No, a cloned voice stays with the provider that trained it, so a switch means cloning again from the original recordings, with the speaker's consent. [How to move a cloned voice to Bland Speech](/blog/move-cloned-voice-to-bland-speech) covers the steps for Bland.
