# Bland Speech vs Azure custom voice cloning (2026): Access, consent, and hosting costs compared

A developer can't sign up for Azure custom voice with a credit card. Microsoft treats it as a Limited Access feature, available by registration only and for approved use cases, and Microsoft says only customers working directly with a Microsoft account team are eligible. For a team without that relationship, the question is settled before audio, price, or voice quality comes up.

Azure also bills for custom voice differently from a stock voice. A custom voice pays $24 per million characters for synthesis, plus $52 for each compute hour of training and $4.04 per hour for each hosted model. Bland Speech publishes one standard rate, $0.04 per 1,000 characters. Bland's [text to speech comparison hub](/compare/text-to-speech) and [voice cloning comparisons](/compare/text-to-speech/voice-cloning) cover the wider field.

## Which is better, Bland Speech or Azure, for voice cloning?

For a team that wants to clone a voice without applying through a Microsoft account team, Bland Speech is the better fit as of November 2026, and Bland Speech v3 ranked first of 13 AI voices in Design Arena's Audio Realism Bench results of 4 August 2026. Azure is the better fit for a team that already works with Microsoft account managers, needs a personal voice across 90+ languages, or wants Microsoft's recorded-consent process with voiceprint matching.

<div class="cmp-usewhen"><div class="cmp-usewhen-own"><p><strong>Use Bland Speech when</strong></p><ul><li>A team will clone a voice from one clean sample of about 10 seconds, or from 30 minutes or more of audio for a professional clone</li><li>A team will confirm that every cloned voice is its to use, and pay per character at one standard rate</li><li>A professional clone will pass Bland's consent and rights review</li><li>Cloned speech will go straight onto a phone line as mu-law at 8 kHz</li></ul></div><div><p><strong>Use Azure when</strong></p><ul><li>A team will register with Microsoft's Limited Access review and work with a Microsoft account team to qualify</li><li>A team will obtain explicit written permission from voice talent and have the talent record a consent statement</li><li>A team will pay for training by the compute hour and for hosting by the model per hour</li><li>A product needs personal voice in more than 90 languages</li></ul></div></div>

| Feature | Bland Speech | Azure |
|---|---|---|
| Audio Realism Bench, 4 August 2026 | Ranked 1st of 13 AI voices, 1365 | Microsoft's MAI-Voice-2 ranked 2nd, 1224 |
| Audio for an instant clone | One clean sample, about 10 seconds | Personal voice: a 5 to 90 second sample plus a spoken consent statement |
| Audio for a professional clone | 30 minutes or more | Custom neural voice lite: 20 to 50 samples recorded in Speech Studio |
| Who can get cloning | The customer confirms each voice is its to use | Limited Access, by registration, for customers managed by Microsoft |
| Consent check | Consent and rights review for professional clones | Recorded consent statement, checked by speaker recognition |
| Price per million characters | $40, one standard rate | $24 custom voice, $48 custom neural HD, plus training and hosting |
| Phone audio | Mu-law at 8 kHz | Mu-law and A-law at 8 kHz |

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

- **Audio Realism Bench, 4 August 2026:** Design Arena's blind listening test, where listeners hear two clips of the same line from two models and pick the one that sounds more like a real person. Microsoft's MAI-Voice-2 was in the results. Microsoft offers MAI-Voice-2.1 and MAI-Voice-2.1-Flash through Azure Speech, and neither was scored in them. Azure's custom and personal voices weren't in the results either. Each model was scored on its stock voices, so the numbers describe a model, not any one clone.
- **Audio for a clone:** Bland says an instant clone needs one clean sample of about 10 seconds and a professional clone is fine-tuned on 30 minutes or more. Microsoft says personal voice needs a 5 to 90 second human speech sample plus a verbal consent statement, and approved registrants of custom neural voice lite record 20 to 50 samples. Microsoft says MAI-Voice instant cloning takes 5 to 60 seconds of audio, by gated access.
- **Who can get cloning:** Microsoft says custom voice is available by registration only, limited to the use cases a customer selects and Microsoft approves, and that only customers managed by Microsoft are eligible. Personal voice also needs approved access.
- **Consent check:** Microsoft requires explicit written permission from voice talent and a recorded acknowledgment statement, which it checks against the training recording with speaker recognition. Bland asks the customer to confirm every voice is theirs to use and reviews professional clones for consent and rights.
- **Price per million characters:** Bland Speech costs $0.04 per 1,000 characters, which is $40 per million. Azure's custom voice synthesis costs $24 per million characters and $48 for neural HD, and training and hosting are billed on top.

</details>

## How do Bland Speech and Azure voice cloning compare?

Microsoft and Bland both put a gate in front of professional cloning, and the gates check different things. Microsoft checks who the customer is and what the voice will be used for. Bland checks that the voice being cloned is the customer's to use.

### Who can get a cloned voice?

Azure custom voice needs registration and approval of the use cases a customer selects, and Microsoft says only customers managed by Microsoft are eligible, meaning those working directly with Microsoft account teams. Personal voice, the short-sample option, also needs approved access. A team outside that relationship can't start a custom voice on its own.

On Bland Speech the customer confirms every voice is theirs to use, and a professional clone passes Bland's consent and rights review before training.

### How much audio does each need?

An instant clone needs about 10 seconds on Bland Speech. On Azure, personal voice takes a human speech sample of 5 to 90 seconds plus a verbal consent statement, and MAI-Voice instant cloning takes 5 to 60 seconds through gated access. For a professional voice, Bland Speech asks for 30 minutes or more. Microsoft's custom neural voice lite, a preview, has approved registrants record 20 to 50 samples in Speech Studio, and Microsoft says lite voices and their recordings delete automatically within 90 days unless full access is granted.

### What does each do to confirm the voice is theirs?

Microsoft publishes the more detailed consent process. Customers must warrant that they obtained explicit written permission from voice talent before creating a model, and the talent records a consent statement. Microsoft says it may use speaker recognition on that recording to confirm the voice signature matches the training data with reasonable confidence. Bland describes a customer's confirmation and a rights review, with every clone private to the customer's organization.

### Bland Speech cloning overview

Bland Speech is the text to speech model inside Bland, a voice AI provider that builds and runs its own speech model for enterprise phone calls. Bland's default cloning engine is BTTS V3, which Bland describes as its highest-fidelity engine, and [Bland Speech](/speech) has a studio for trying a clone. Bland's [text to speech API](/text-to-speech-api) takes POST /v2/tts for complete text and WSS /v2/tts/ws for streamed text.

<div class="cmp-proscons cmp-proscons-own"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>An instant clone from one clean sample of about 10 seconds</li><li>A professional clone fine-tuned on 30 minutes or more of audio, after a consent and rights review</li><li>Every clone private to the customer's organization</li><li>No monthly plan, with one standard rate of $0.04 per 1,000 characters</li><li>Mu-law at 8 kHz and PCM from 8 to 48 kHz for phone lines</li><li>Bland Speech v3 ranked first of 13 AI voices on Audio Realism Bench on 4 August 2026</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>

### Azure custom voice overview

Azure Speech is Microsoft's speech service, with standard neural voices in 100+ languages and locales, HD voices, MAI-Voice models in public preview, personal voice, and custom professional voice. Microsoft says MAI-Voice is in public preview without a service-level agreement and isn't recommended for production workloads.

<div class="cmp-proscons" data-rival="azure"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Recorded consent from voice talent, checked by speaker recognition</li><li>Personal voice in more than 90 languages across 100+ locales</li><li>Mu-law and A-law output at 8 kHz</li><li>A free tier of 0.5 million neural characters a month on standard voices</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Custom voice is Limited Access and open to customers managed by Microsoft</li><li>Training costs $52 per compute hour and hosting $4.04 per model per hour</li><li>MAI-Voice, which includes instant cloning, is in public preview with no service-level agreement</li><li>Custom neural voice lite recordings delete within 90 days unless full access is granted</li></ul></div></div>

## How much does a cloned voice cost on Bland Speech and Azure?

An Azure custom voice costs less per character than Bland Speech and more per month, because Azure bills hosting by the hour on top of synthesis. A stock Azure Neural voice at $15 per million characters is cheaper than Bland Speech at $40, and a custom voice at $24 per million is cheaper too, before training and hosting.

### Bland Speech pricing

Bland Speech costs $0.04 per 1,000 characters in the studio or the API, with no monthly plan. New accounts get 50,000 characters free, and credits don't expire. Enterprise pricing is custom, set with Bland's sales team.

### Azure custom voice pricing

Azure charges $24 per million characters for custom voice synthesis and $48 for neural HD, $52 per compute hour for voice model training up to $936 per training, and $4.04 per model per hour for endpoint hosting. Microsoft's commitment tiers for neural voices exclude custom neural voice and personal voice. Microsoft says prices are estimates and vary by region.

### What a month of cloned speech costs

Each Azure figure adds synthesis to 720 hours of endpoint hosting, which is a 30-day month with the endpoint left running, and leaves out the one-time training charge of up to $936.

- **100,000 characters a month:** Bland Speech $4, Azure about $2,911 ($2.40 of synthesis plus $2,909 of hosting).
- **1 million characters a month:** Bland Speech $40, Azure about $2,933 ($24 plus $2,909).
- **10 million characters a month:** Bland Speech $400, Azure about $3,149 ($240 plus $2,909).

## Which sounds more human, Bland Speech or Azure?

Bland Speech v3 scored 1365 in the blind tests Design Arena published on 4 August 2026, 141 points above Microsoft's MAI-Voice-2 at 1224, and ranked first of 13 AI voices in those results, with MAI-Voice-2 2nd.

Listeners hear two clips of the same line without knowing which model made which and pick the one that sounds more like a real person. Azure's custom and personal voices weren't in those results, and neither were the MAI-Voice-2.1 models that Azure Speech offers. Design Arena scored one female and one male stock voice per model in its [published Audio Realism Bench results](https://notes.designarena.ai/audio-realism-benchmark-measuring-realism-in-audio-models/), so the result describes how a model sounds and a particular clone can land above or below it.

## Why is Bland Speech the best Azure custom voice alternative for phone calls?

A team picks Bland Speech over Azure custom voice for phone calls when it wants to start a clone without an account-team relationship, wants one standard per-character rate in place of training and hosting meters, and wants a voice that ranked first among AI voices on Audio Realism Bench on 4 August 2026. Both return mu-law at 8 kHz, so the phone format isn't the difference.

One more fact matters to a regulated team. Bland is SOC 2 Type II audited, HIPAA compliant with a signed BAA, and FedRAMP 20x Class C certified at Moderate, granted on 17 November 2026, and Bland says speech synthesis runs inside the FedRAMP boundary.

A voice cloned on Azure doesn't move across. A team clones it again on Bland Speech from the original recordings with the voice owner's consent, and the [guide to moving a cloned voice to Bland Speech](/blog/move-cloned-voice-to-bland-speech) covers the steps. A team weighing Azure's whole speech service can read [Azure text to speech alternatives](/blog/azure-text-to-speech-alternatives) or [Bland Speech vs Azure text to speech](/compare/bland-speech-vs-azure-tts).

## Frequently asked questions

### Is Bland Speech better than Azure for voice cloning?

For a team without a Microsoft account team, yes. Azure custom voice is Limited Access and open to customers managed by Microsoft, while Bland Speech takes a customer's confirmation that each voice is theirs to use, and Bland Speech v3 ranked first of 13 AI voices on Audio Realism Bench on 4 August 2026.

### Can anyone sign up for Azure custom voice?

No. Microsoft says custom neural voice is a Limited Access feature available by registration only and for approved use cases, and that only customers managed by Microsoft are eligible.

### How much audio does Azure need to clone a voice?

Personal voice needs a 5 to 90 second human speech sample plus a verbal consent statement, and MAI-Voice instant cloning takes 5 to 60 seconds. Custom neural voice lite has approved registrants record 20 to 50 samples.

### How much does an Azure custom voice cost?

Synthesis costs $24 per million characters. Training costs $52 per compute hour up to $936, and hosting costs $4.04 per model per hour, which is about $97 a day if the endpoint stays up.

### How much audio does Bland Speech need to clone a voice?

An instant clone needs one clean sample of about 10 seconds. A professional clone is fine-tuned on 30 minutes or more of audio and passes Bland's consent and rights review.

### What does Bland Speech charge for a cloned voice?

Bland publishes one standard rate for Bland Speech, $0.04 per 1,000 characters in the studio or the API, with no monthly plan.

### Can an Azure clone move to Bland Speech?

No file moves across. A team clones the voice again on Bland Speech from the original recordings and confirms it is the team's to use.
