A phone call carries audio at 8 kHz, one sixth of the 48 kHz Bland Speech renders natively, so the voice a caller hears is a clone after a conversion built for telephony. A cloned voice for a phone line should be judged after that conversion, and the service behind it has to hand over audio in the format the line takes.
Six services make clones and return 8 kHz mu-law, and they differ in how much audio a clone needs, what each company says about first audio, and which telephony tools each one plugs into.
What is the best voice cloning for phone calls?
Bland Speech is the best voice cloning for phone calls, as of November 2026, because Bland Speech returns 8-bit mu-law at 8 kHz, a new context on its WebSocket stops the turn that is playing when a caller cuts in, and Bland Speech v3 ranked first of 13 AI voices on Design Arena's Audio Realism Bench results of 4 August 2026, a test with phone-agent turns among its prompts.
That ranking rests on phone output and the realism result, not on speed or price, and Cartesia says first audio arrives in under 90ms against the 199ms Bland publishes. Design Arena's test used one female and one male American-English voice per model, so it scores a model's own voices and not a clone.
- Bland Speech: Best overall for a cloned voice on a phone line
- Cartesia: Best for documented telephony formats and a published sub-90ms claim
- Rime: Best for regulated telephony, with clones ordered from Rime
- ElevenLabs: Best for clones on Eleven v4 with mu-law and A-law output
- Inworld: Best for a clone from 3 seconds that returns mu-law and A-law
- MiniMax: Best for paying per cloned voice, with G.711 mu-law output
What these rows mean
- 8 kHz mu-law: the audio format US phone networks carry, as each service's own format list shows.
- First audio: each company's own published claim, measured its own way, and Cartesia, ElevenLabs, and Inworld state theirs as time to first audio or median latency excluding network time. None is a test Bland ran.
- Clone from: the audio each service documents for an instant clone, then for a professional one where it has both.
- Telephony tools: the agent and telephony platforms each service names, plus the services Pipecat and LiveKit support.
What should a team check before putting a cloned voice on a phone line?
A team should check the format first, because a service that can't return mu-law leaves the team converting every call's audio. Bland Speech offers 8-bit mu-law at 8 kHz and PCM at 8, 16, 24, 44.1, and 48 kHz, and the other five also return mu-law at 8 kHz. A clone should be heard in the phone format, not only in a studio.
The second check is the path into the call. All six work with Pipecat or LiveKit Agents, two voice agent frameworks, and Bland Speech has a Pipecat service maintained by the Pipecat team and a LiveKit Agents plugin for Python. Bland's text to speech API page has the endpoints.
The third check is interruption. A caller who talks over a cloned voice should stop it, and on Bland Speech a new context ID on the WebSocket preempts the turn that is playing. MiniMax's bidirectional WebSocket takes a cancel message for the same job.
How do the 6 best voice cloning services for phone calls compare?
These six differ most in documented telephony support and in how a clone gets made, and Bland Speech is placed first because its clones return mu-law and handle interruptions, from the company whose Bland Speech v3 voices ranked first in Design Arena's tests of 4 August 2026.
1. Bland Speech: Best overall for a cloned voice on a phone line
An instant clone on Bland Speech needs one clean sample of about 10 seconds, and a professional clone is fine-tuned on 30 minutes or more of audio and goes through Bland's consent and rights review. Every clone is private to the customer's organization, and the customer confirms the voice is theirs to use. Bland Speech streams over POST /v2/tts and WSS /v2/tts/ws, which return PCM from 8 to 48 kHz or 8 kHz mu-law, and the WebSocket takes language model tokens as they arrive.
Bland Speech has a Pipecat service maintained by the Pipecat team and a LiveKit Agents plugin for Python.
Pros
- 8 kHz mu-law and PCM up to 48 kHz
- Bland publishes 199ms to first audio
- A new context on the WebSocket interrupts the turn that is playing
- A Pipecat service and a LiveKit plugin
- Bland Speech v3 ranked first of 13 AI voices on Audio Realism Bench on 4 August 2026
Cons
- More than a hobby project needs, because Bland Speech is built for production calls
Pricing: $0.04 per 1,000 characters, which is $40 per million, pay as you go with no monthly plan, and 50,000 characters free once for a new account. The Bland Speech page has a studio to try a clone.
2. Cartesia: Best for documented telephony formats and a published sub-90ms claim
Cartesia names mu-law at 8 kHz for North American and Japanese telephony, and offers A-law as well. Instant clones need 10 seconds of audio, and professional cloning is on the $49 Startup plan and above. Cartesia also sells Managed Agents, which run a language model for the customer, with Cartesia phone numbers at $0.014 a minute. Bland Speech vs Cartesia voice cloning compares the two side by side.
Pros
- Mu-law and A-law at 8 kHz, documented for telephony
- Cartesia says first audio arrives in under 90ms
- Training a Pro Voice Clone is free
Cons
- Professional cloning starts on the $49 Startup plan
- Speed and volume don't apply to Pro Voice Clones
Pricing: 1 credit per character from monthly plans: Pro at $5 for 100,000 credits, Startup at $49 for 1.25 million, and Scale at $299 for 8 million, with overages at $65, $45, or $38 per million credits.
3. Rime: Best for regulated telephony, with clones ordered from Rime
Rime says its voice models are made for production voice agents, IVR, and telephony in healthcare, financial services, and other regulated industries. Rime's cloud API returns G.711 mu-law, and Rime sets a sampling rate of 8000 to match a telephony stream. Rime sells enterprise voice cloning by order on Coda or Mist, from 30 to 60 minutes of clean recordings for a basic clone and two to five hours or more for the strongest results.
Rime is SOC 2 Type 2 compliant, HIPAA compliant with a BAA available, and says enterprise customers can run it in a dedicated VPC or fully on-premises.
Pros
- Built for telephony, with mu-law at 8 kHz
- SOC 2 Type 2 and HIPAA compliance, with a BAA available
- LiveKit, Pipecat, Daily, and Vapi guides
Cons
- Voice cloning is an enterprise order placed with Rime
- Coda's cloud API covers 10 languages, with each voice serving one
- Priced by the minute rather than by the character
Pricing: $0.030 a minute for Mist v3 and $0.050 for Coda v2 on Flex, about 1,000 characters a minute, with $50 of usage free each month and custom enterprise pricing.
4. ElevenLabs: Best for clones on Eleven v4 with mu-law and A-law output
ElevenLabs returns mu-law and A-law at 8 kHz and works with Pipecat and LiveKit Agents. Instant cloning starts on its $6 Starter plan, from 10 seconds on Eleven v4, and professional cloning on its $22 Creator plan, from about 30 minutes.
Pros
- Instant clones from 10 seconds on Eleven v4
- Mu-law and A-law output at 8 kHz
- A Pipecat service and a LiveKit plugin for Python and Node.js
Cons
- Cloning needs the $6 Starter or $22 Creator plan
- The standard text to speech WebSocket doesn't accept Eleven v3 or v4
- Eleven v4 costs $0.08 per 1,000 characters
Pricing: $0.08 per 1,000 characters for Eleven v4, v3, and Multilingual v2, and $0.04 for v4 Turbo, v3 Conversational, and Flash, against plans from Free to Business at $990 a month.
5. Inworld: Best for a clone from 3 seconds that returns mu-law and A-law
Inworld clones from as little as 3 seconds of audio on every plan, and a professional clone, in beta on TTS-2, needs at least 10 minutes. It returns mu-law and A-law at 8 kHz, says first audio arrives as low as 50 ms, and has integrations for Pipecat, LiveKit, and Vapi. Inworld says cloned voices bill at the stock-voice rate.
Pros
- Instant cloning from 3 seconds on every plan
- Mu-law and A-law output at 8 kHz
- Pipecat, LiveKit, and Vapi integrations
- Cloned voices bill at the stock-voice rate
Cons
- Zero data retention doesn't cover clone samples
- Professional cloning is in beta and runs on TTS-2 only
Pricing: $25 per million characters for TTS-2 and $15 for TTS-2 Flash on On-Demand pay as you go, with lower rates on monthly plans from Creator at $25.
6. MiniMax: Best for paying per cloned voice, with G.711 mu-law output
MiniMax returns G.711 mu-law at 8 kHz as pcmu_raw or pcmu_wav, and its bidirectional WebSocket accepts language model output at any granularity and takes a cancel message for interruptions. A clone needs 10 seconds to 5 minutes of audio, costs $1.5 per voice, and is deleted if it isn't used within 7 days. MiniMax has a LiveKit Agents plugin.
Pros
- G.711 mu-law output at 8 kHz
- A bidirectional WebSocket with a cancel message
- $1.5 per cloned voice, billed at first use
Cons
- A cloned voice is deleted after 7 days without use
- Speech costs $60 to $100 per million characters
Pricing: $100 per million characters for speech-2.8-hd and $60 for speech-2.8-turbo on pay as you go, plus $1.5 per cloned voice.
How were these voice cloning services compared for phone calls?
Samuel Oram-Jones compared six services on phone formats, each company's published first-audio claim, the audio a clone needs, and the telephony tools each names. Rival details are what each company publishes about itself, and Bland Speech sits first on format support and its realism result, not on a speed figure or price.
Bland's comparison of text to speech APIs sorts these companies by job, and its voice cloning comparisons take one provider at a time.
Frequently asked questions
What is the best voice cloning for phone calls?
Bland Speech, for a cloned voice that has to return 8 kHz mu-law and stop when a caller cuts in. Bland publishes 199ms to first audio and Cartesia says under 90ms, so the pick rests on phone output and on Bland Speech v3 ranking first of 13 AI voices on 4 August 2026, not on speed.
What audio format does a phone call need?
US phone networks carry mu-law at 8 kHz. Bland Speech, Cartesia, ElevenLabs, Rime, Inworld, and MiniMax all return it, and Cartesia, ElevenLabs, and Inworld return A-law as well. Bland's guide to mu-law against PCM explains the formats.
Does a cloned voice sound different on a phone line?
A phone line keeps less of a recording than a studio file does, because 8 kHz mu-law has one sixth of the samples of 48 kHz audio at 8 bits each, so a clone should be heard after that conversion before it ships.
How fast does a cloned voice start speaking?
Each company publishes its own figure, and none is a race Bland ran. Cartesia says under 90ms, ElevenLabs about 100ms median on v4 Turbo excluding network time, Inworld as low as 50 ms, and Bland Speech 199ms to first audio on a streamed request.
How much audio does a clone for a phone agent need?
An instant clone needs 3 seconds at Inworld and about 10 seconds at Bland Speech, Cartesia, and ElevenLabs, and a professional clone is fine-tuned on 30 minutes or more at Bland Speech and needs about 30 minutes at ElevenLabs. Bland's guide to how much audio a clone needs covers sample quality.
What does a cloned phone voice cost to run?
Bland Speech costs $40 per million characters with no monthly plan, Inworld $15 to $25, ElevenLabs $40 to $80, MiniMax $60 to $100 plus $1.5 per voice, and Rime $0.03 to $0.05 a minute. Cartesia charges 1 credit per character from monthly plans.
