How we rate AI audio tools

Every tool starts at zero and earns points for what it does and what it costs. Here is exactly what earns them.

Last updated September 2026

The short version

These tools do different jobs, so each kind has its own checklist: AI voice generators on 4 things, speech-to-text APIs on 4 things, AI transcription tools on 4 things, audio cleanup tools on 4 things and AI music generators on 4 things. Each checklist is worth 100 points, and a tool is only ranked against its own kind.

The final score is those points out of 10. A tool with 84 points scores 8.4.

Everything comes from the tool's own website. We don't take payment for points.

AI voice generators

What you can makeWorkflow and outputPricingPrivacy and trust

What earns points

1. What you can make

40 of 100 points

What you can make: cloning a voice from your own recording, how many languages and ready-made voices, dubbing into another language, changing the voice of a recording, a long-form editor for audiobooks, conversations between several voices, reading your documents aloud, and fixing how a word is said.

Clones a voice from your recordingup to 2

Yes	+2
No	0

Languages it speaksup to 1.5

20 languages	+0.5
21 to 40 languages	+1
41 languages or more	+1.5

Ready-made voicesup to 1.5

250 voices	+0.5
251 to 800 voices	+1
801 voices or more	+1.5

Dubs speech into another languageup to 1

Yes	+1
No	0

Changes the voice of a recordingup to 1

Yes	+1
No	0

Editor for audiobooks and long projectsup to 1

Yes	+1
No	0

Conversations between several voicesup to 1

Yes	+1
No	0

Reads your documents aloudup to 0.5

Yes	+0.5
No	0

Fixes how a word is saidup to 0.5

Yes	+0.5
No	0

Each section adds up to 10, then counts for 40 of the 100.

2. Workflow and output

20 of 100 points

The audio formats it exports, where it runs, the apps it connects to, and a shared team workspace.

File formats it exportsup to 2.5

1 export format	+1
2 export formats	+1.5
3 export formats or more	+2.5

Where it runsup to 2.5

1 platform	+0.5
2 platforms	+1.5
3 platforms or more	+2.5

Apps it connects toup to 2.5

None	0
1 to 3 integrations	+1.5
4 integrations or more	+2.5

Shared team workspaceup to 2.5

Yes	+2.5
No	0

Each section adds up to 10, then counts for 20 of the 100.

3. Pricing

22 of 100 points

What the cheapest paid plan costs each month, and whether there is a free plan.

Cheapest paid plan, monthlyup to 7.5

Up to $6	+7.5
Up to $10	+6
Up to $20	+4.5
Up to $30	+3
Over $30	+1.5

Free planup to 2.5

Yes	+2.5
No	0

Each section adds up to 10, then counts for 22 of the 100.

4. Privacy and trust

18 of 100 points

Does it train its AI on what you upload, does it hold SOC 2 and state GDPR, and how many ways are there to reach support?

Does not train AI on what you uploadup to 4

Yes	+4
Only if you opt out	+2
No	0

SOC 2up to 2

Yes	+2
No	0

GDPRup to 1.5

Yes	+1.5
No	0

Ways to reach supportup to 2.5

1 support channel	+0.5
2 support channels	+1
3 support channels or more	+2.5

Each section adds up to 10, then counts for 18 of the 100.

A real example: ElevenLabs

Here is how ElevenLabs got its 8.5.

Section	Points
What you can make	40 of 40
Workflow and output	11 of 20
Pricing	22 of 22
Privacy and trust	11.7 of 18
Total	84.7 of 100 = 8.5 out of 10

See every fact behind ElevenLabs's score.

Speech-to-text APIs

What the API doesOutput and integrationsPricingPrivacy and trust

What earns points

1. What the API does

40 of 100 points

How many languages it transcribes, running in your own cloud, text to speech on the same platform, redacting personal data and time codes per word. Real-time streaming, speaker labels, custom vocabulary, HIPAA, SOC 2 and GDPR are on all of them, so they earn no points.

Languages it transcribesup to 3

50 languages	+1
51 to 60 languages	+2
61 languages or more	+3

Runs in your own cloud or on-premisesup to 2.5

Yes	+2.5
No	0

Also sells text to speechup to 2

Yes	+2
No	0

Redacts personal dataup to 1.5

Yes	+1.5
No	0

Time codes per wordup to 1

Yes	+1
No	0

Each section adds up to 10, then counts for 40 of the 100.

2. Output and integrations

20 of 100 points

The output formats and the apps it connects to.

Output formatsup to 5

1 format	+2
2 to 3 formats	+4
4 formats or more	+5

Apps it connects toup to 5

None	0
1 integration	+2.5
2 integrations or more	+5

Each section adds up to 10, then counts for 20 of the 100.

3. Pricing

22 of 100 points

The cheapest pay-as-you-go rate for transcribing an hour of recorded audio.

Cheapest pre-recorded transcription, per hourup to 10

Up to $0.15	+10
Up to $0.20	+8
Up to $0.30	+6
Over $0.30	+3

Each section adds up to 10, then counts for 22 of the 100.

4. Privacy and trust

18 of 100 points

Does it train its models on your audio, and how many ways are there to reach support?

Does not train AI on what you uploadup to 5

Yes	+5
Only if you opt out	+2.5
No	0

Ways to reach supportup to 5

1 support channel	+1
2 support channels	+2
3 support channels or more	+5

Each section adds up to 10, then counts for 18 of the 100.

A real example: AssemblyAI

Here is how AssemblyAI got its 8.6.

Section	Points
What the API does	40 of 40
Output and integrations	10 of 20
Pricing	22 of 22
Privacy and trust	13.5 of 18
Total	85.5 of 100 = 8.6 out of 10

See every fact behind AssemblyAI's score.

AI transcription tools

What you get from a recordingWorkflow and outputPricingPrivacy and trust

What earns points

1. What you get from a recording

40 of 100 points

What you get from a recording: how many languages, an editor synced to the audio, show notes and posts written from it, translation, summaries, clips, live transcription and transcribing on your own device.

Languages it transcribesup to 2

25 languages	+0.5
26 to 50 languages	+1.5
51 languages or more	+2

Editor synced to the audioup to 1.5

Yes	+1.5
No	0

Writes show notes, posts and articlesup to 1.5

Yes	+1.5
No	0

Translates transcriptsup to 1

Yes	+1
No	0

Summaries and chaptersup to 1

Yes	+1
No	0

Cuts shareable clipsup to 1

Yes	+1
No	0

Transcribes liveup to 1

Yes	+1
No	0

Transcribes on your own deviceup to 1

Yes	+1
No	0

Each section adds up to 10, then counts for 40 of the 100.

2. Workflow and output

20 of 100 points

The file formats it exports, the apps it connects to, where it runs, a team workspace and an API.

File formats it exportsup to 3

2 export formats	+1
3 to 4 export formats	+2
5 export formats or more	+3

Apps it connects toup to 3

None	0
1 to 5 integrations	+1.5
6 integrations or more	+3

Where it runsup to 2

1 platform	+0.5
2 platforms	+1.5
3 platforms or more	+2

Shared team workspaceup to 1

Yes	+1
No	0

APIup to 1

Yes	+1
No	0

Each section adds up to 10, then counts for 20 of the 100.

3. Pricing

22 of 100 points

What the cheapest paid plan costs each month, and whether there is a free plan.

Cheapest paid plan, monthlyup to 7

Up to $15	+7
Up to $20	+5.5
Up to $30	+4
Over $30	+2

Free planup to 3

Yes	+3
No	0

Each section adds up to 10, then counts for 22 of the 100.

4. Privacy and trust

18 of 100 points

Does it train AI on your recordings, does it hold SOC 2, state GDPR, offer HIPAA and single sign-on, and how many ways are there to reach support?

Does not train AI on what you uploadup to 3.5

Yes	+3.5
Only if you opt out	+1.8
No	0

SOC 2up to 1.5

Yes	+1.5
No	0

GDPRup to 1

Yes	+1
No	0

HIPAAup to 1.5

Yes	+1.5
No	0

Single sign-onup to 1

Yes	+1
No	0

Ways to reach supportup to 1.5

1 support channel	+0.3
2 support channels	+0.6
3 support channels or more	+1.5

Each section adds up to 10, then counts for 18 of the 100.

A real example: Sonix

Here is how Sonix got its 6.6.

Section	Points
What you get from a recording	22 of 40
Workflow and output	17 of 20
Pricing	8.8 of 22
Privacy and trust	18 of 18
Total	65.8 of 100 = 6.6 out of 10

See every fact behind Sonix's score.

Audio cleanup tools

What it fixesWorkflow and outputPricingPrivacy and trust

What earns points

1. What it fixes

40 of 100 points

What it fixes beyond noise, which every tool here removes: rebuilding speech to sound studio-recorded, setting loudness to a target, cutting filler words and silence, processing several tracks together, working live on calls, batch processing and publishing straight to a host.

Rebuilds speech to sound studio-recordedup to 1.5

Yes	+1.5
No	0

Sets loudness to a targetup to 1.5

Yes	+1.5
No	0

Cuts filler wordsup to 1.5

Yes	+1.5
No	0

Processes several tracks togetherup to 1.5

Yes	+1.5
No	0

Cuts silenceup to 1

Yes	+1
No	0

Works live on callsup to 1

Yes	+1
No	0

Processes many files at onceup to 1

Yes	+1
No	0

Publishes straight to a hostup to 1

Yes	+1
No	0

Each section adds up to 10, then counts for 40 of the 100.

2. Workflow and output

20 of 100 points

The file formats it exports, the apps it connects to, where it runs, running as a plugin inside your editor, and an API.

File formats it exportsup to 2.5

None	0
1 to 2 export formats	+1
3 to 6 export formats	+2
7 export formats or more	+2.5

Apps it connects toup to 2.5

4 integrations	+1
5 to 10 integrations	+2
11 integrations or more	+2.5

Where it runsup to 2

1 platform	+0.5
2 platforms	+1.2
3 platforms or more	+2

Runs as a plugin in your editorup to 1.5

None	0
1 plugin formats or more	+1.5

APIup to 1.5

Yes	+1.5
No	0

Each section adds up to 10, then counts for 20 of the 100.

3. Pricing

22 of 100 points

What the cheapest paid plan costs each month, whether there is a free plan, and whether you can buy a licence you own outright.

Cheapest paid plan, monthlyup to 6

Up to $12	+6
Up to $16	+4.5
Up to $20	+3
Over $20	+1.5

Free planup to 2

Yes	+2
No	0

Sells a licence you own outrightup to 2

Yes	+2
No	0

Each section adds up to 10, then counts for 22 of the 100.

4. Privacy and trust

18 of 100 points

Does it train AI on your audio, does it hold SOC 2 and state GDPR, and how many ways are there to reach support? Software you install and run yourself is not asked the first three.

Does not train AI on what you uploadup to 4

Yes	+4
Only if you opt out	+2
No	0

SOC 2up to 2

Yes	+2
No	0

GDPRup to 1.5

Yes	+1.5
No	0

Ways to reach supportup to 2.5

1 support channel	+0.5
2 support channels	+1
3 support channels or more	+2.5

Each section adds up to 10, then counts for 18 of the 100.

A real example: Auphonic

Here is how Auphonic got its 7.2.

Section	Points
What it fixes	36 of 40
Workflow and output	14 of 20
Pricing	14.3 of 22
Privacy and trust	8.1 of 18
Total	72.4 of 100 = 7.2 out of 10

See every fact behind Auphonic's score.

AI music generators

What you can makeWorkflow and outputPricing and rightsPrivacy and trust

What earns points

1. What you can make

40 of 100 points

What you can make: songs with vocals, separate stems, editing or extending part of a track, singing your own lyrics, starting from audio you upload, and writing lyrics for you. All of them make instrumentals, so that earns no points.

Songs with sung vocalsup to 2

Yes	+2
No	0

Downloads separate stemsup to 2

Yes	+2
No	0

Edits or extends part of a trackup to 2

Yes	+2
No	0

Sings lyrics you writeup to 1.5

Yes	+1.5
No	0

Starts from audio you uploadup to 1.5

Yes	+1.5
No	0

Writes lyrics for youup to 1

Yes	+1
No	0

Each section adds up to 10, then counts for 40 of the 100.

2. Workflow and output

20 of 100 points

The formats you can download, where it runs, an API and the apps it connects to.

File formats you can downloadup to 3

None	0
1 format	+1.5
2 formats or more	+3

Where it runsup to 3

1 platform	+1
2 platforms	+2
3 platforms or more	+3

APIup to 2

Yes	+2
No	0

Apps it connects toup to 2

None	0
1 integrations or more	+2

Each section adds up to 10, then counts for 20 of the 100.

3. Pricing and rights

22 of 100 points

What the cheapest paid plan costs each month, whether a paid plan lets you use songs commercially, and whether you own them.

Cheapest paid plan, monthlyup to 4

Up to $8	+4
Up to $10	+3
Over $10	+2

Commercial use on a paid planup to 3

Yes	+3
No	0

You own what you make on a paid planup to 3

Yes	+3
No	0

Each section adds up to 10, then counts for 22 of the 100.

4. Privacy and trust

18 of 100 points

Does it train its AI on what you upload, does it state GDPR, and how many ways are there to reach support?

Does not train AI on what you uploadup to 4.5

Yes	+4.5
Only if you opt out	+2.3
No	0

GDPRup to 2.5

Yes	+2.5
No	0

Ways to reach supportup to 3

1 support channel	+0.6
2 support channels	+1.2
3 support channels or more	+3

Each section adds up to 10, then counts for 18 of the 100.

A real example: Suno

Here is how Suno got its 8.4.

Section	Points
What you can make	40 of 40
Workflow and output	16 of 20
Pricing and rights	22 of 22
Privacy and trust	5.6 of 18
Total	83.6 of 100 = 8.4 out of 10

See every fact behind Suno's score.

When a tool doesn't say

If a tool's website doesn't answer a yes-or-no question, it gets half the points for that item. Not zero, because a missing answer isn't the same as a no, and not full marks, because silence shouldn't beat a vendor who answers honestly. For a number, like a price, it gets the typical value of the tools that do publish one. Every item scored this way is marked on the tool's scorecard.

If a tool leaves out more than a third of what we check, we don't show a score for it at all. A number built mostly on guesses wouldn't be fair to you or to the tool. Right now 24 of the 24 tools we checked have enough information to be scored.

What the score doesn't tell you

We score what a tool offers and what it costs, as its own website describes it. That's not a hands-on test. The score can't tell you whether a tool is easy to use, whether its data is accurate, or whether support answers quickly. The review covers that.

See the AI audio tools ranked or read how we review tools in general.