Picovoice Leopard
picovoice.ai
- Free trial
- Windows
- macOS
- Linux
- Web
- iPhone and iPad
- Android
- API
- Self-hosted
What Picovoice Leopard is
Picovoice Leopard is an on-device speech-to-text engine for transcribing audio and video files such as interviews, meetings, calls, podcasts and voicemails. Processing stays locally on the device; Picovoice says inputs and outputs are not transmitted to, stored by or accessible to it or its service providers. Transcripts can include each word’s start and end times, confidence score and speaker tag. Developers can add custom vocabulary and boost selected words or phrases through Picovoice Console. Supported languages are English, French, German, Italian, Japanese, Korean, Portuguese and Spanish. Platforms include Linux, macOS, Windows, Android, iOS, web and Raspberry Pi, with SDK quick starts across these environments and multiple frameworks. Use requires an AccessKey, and usage is measured by audio processed per second, resetting every 30 days. Leopard does not provide real-time transcription. Pricing is on request; the trial is free for seven days.
Who it suits
It is designed for enterprise and other B2B developers building local transcription into applications. It is not offered with dedicated personal or non-commercial plans.
Good to know, and what to watch
- Processes speech locally on-device
- Provides word timestamps and speaker tags
- Supports custom vocabulary
- Offers multiple platforms and SDKs
- Seven-day trial needs no credit card
- Does not support real-time transcription
- Requires a Picovoice AccessKey
- Pricing is available on request
- No dedicated personal-use plans
Plans
Every plan and priceFree Trial
Pricing on request · 7-day trial
Getting started
- Open the Picovoice Leopard website
- Choose a supported SDK, API, or platform integration
- Create or obtain a Picovoice AccessKey
- Add the key to your application
- Process recorded audio or video within your account limits
Limits to check
The free trial lasts seven days and includes access to all Picovoice products. Ongoing usage is tracked by audio processed per second and resets every 30 days; account limits are checked through the AccessKey.
The verdict
Leopard focuses on private, local transcription with detailed word metadata and broad SDK coverage. Confirm that file-based processing, AccessKey usage limits and enterprise-oriented terms fit your application.
Questions people ask
Is there a free plan?
There is no free plan. A Free Trial costs nothing, lasts seven days, requires no credit card, and does not auto-renew.
What platforms does Leopard support?
Listed platforms include Android, API, iOS, Linux, macOS, self-hosted, web, Windows, and Raspberry Pi.
Does Leopard work in real time?
No. Leopard does not support real-time transcription; Picovoice directs streaming transcription users to Cheetah.
Which languages are listed?
English, French, German, Italian, Japanese, Korean, Portuguese, and Spanish.
Does Leopard provide speaker labels and timestamps?
Yes. Transcriptions can include speaker tags plus each word's start time, end time, and confidence score.
How is Leopard priced?
Pricing is on request. Use requires a Picovoice AccessKey, and usage is measured by audio processed per second.
What the maker publishes
- Access and usage
- Using Picovoice requires an AccessKey, which authenticates and authorizes use and verifies that usage is within account limits. source
- Access requirement
- Using Leopard requires a Picovoice AccessKey, which authenticates use and verifies it is within account limits. source
- API access
- Yes source
- Audience
- Picovoice describes itself as a B2B company focused on on-device AI for enterprises and says it has no dedicated plans for personal or non-commercial use. source
- Customization
- Picovoice Console lets developers add custom vocabulary and boost the probability of selected words and phrases. source
- Customization
- Picovoice Console lets users add custom vocabulary, specify IPA pronunciations, and boost frequent words and phrases in speech-to-text models. source
- Data use
- Picovoice says it cannot access, collect, or use end-user data to train or improve its AI models, except for custom development where a customer explicitly provides data under a separate written agreement. source
- Free plan
- No source
- Free trial
- Picovoice offers enterprise developers a one-time free trial with no credit card required; it does not renew automatically. source
- Integrations
- Picovoice provides Leopard SDK quick starts for Android, C, .NET, Flutter, iOS, Java, Linux, macOS, Node.js, Python, Raspberry Pi, React, React Native, web, and Windows. source
- Intended users
- Picovoice describes itself as a B2B company focused on on-device AI tools for enterprises and says it has no dedicated plans for personal or non-commercial use. source
- Languages
- The Leopard SDK introduction lists English, French, German, Italian, Japanese, Korean, Portuguese, and Spanish as supported languages. source
- Languages
- Leopard supports English, French, German, Italian, Japanese, Korean, Portuguese, and Spanish; other languages are available to Enterprise Plan customers through Picovoice contacts. source
- Local processing
- Picovoice says Leopard processes voice locally on the device. source
- Local processing
- Leopard processes voice locally on the device. source
- Purpose
- Leopard is an on-device speech-to-text engine for transcribing audio and video files, including recordings of interviews, meetings, calls, podcasts, and voicemails. source
- Real-time limitation
- Leopard does not support real-time transcription; Picovoice directs users needing streaming transcription to Cheetah. source
- SDK availability
- Leopard SDK quick starts are listed for Android, C, .NET, Flutter, iOS, Java, Linux, macOS, Node.js, Python, Raspberry Pi, React, React Native, Web, and Windows. source
- Security
- Picovoice says end-user inputs and generated outputs remain on-device and are not transmitted to, stored by, or accessible to Picovoice or its service providers. source
- Security and compliance
- Picovoice says it does not list security certifications because it cannot access audio processed by its engines or their outputs. source
- Speaker identification
- Yes source
- Support
- Enterprise customers can customize support, prospects can contact sales for dedicated support, and enterprise developers can report bugs through GitHub issues. source
- Support
- Enterprise customers can customize their support level, prospects can contact sales for dedicated support, and enterprise developers can report code or documentation issues on GitHub. source
- Supported platforms
- The SDK introduction lists Linux, macOS, Windows, Android, iOS, web, and Raspberry Pi as supported platforms. source
Checked against the maker's pages on 10 Oct 2026.