What is Desert Ant Labs
Desert Ant Labs is a frontier AI lab building on-device intelligence. They offer fast, specialized models that beat generalist cloud models on performance, cost, and energy. The models run locally on devices, eliminating token costs and enabling privacy-preserving AI. With a native SDK, these models can be integrated into any app with just a few lines of code.
How to use Desert Ant Labs
- Choose a model from the available list (e.g., Voz for speech recognition, Clear for speech enhancement, Redact for PII redaction).
- Integrate the model using the provided SDK for Swift, Kotlin, or JavaScript.
- Add the model to your app with a few lines of code.
- Run inference on-device with no token costs.
- For detailed instructions, refer to the documentation.
Features of Desert Ant Labs
- On-device AI models for speech, text, and vision tasks.
- Fast and specialized models that outperform generalist cloud models.
- No token costs – inference is unlimited and free up to 100k monthly active devices.
- Native SDKs for Swift, Kotlin, and JavaScript.
- Privacy-focused – data stays on the device.
- Wide range of models including:
- Audio: Align (word timestamps), Clear (speech enhancement), Clips (clip selection), Ear (spoken language detection), Uhm (filler-word detection), Voz (speech recognition), Who (speaker labeling).
- Text: Emo (emoji suggestions), Gist (content topic tagging), Redact (PII redaction), Title (titles and descriptions), Tongue (language identification), Schemer (structured extraction), Toxic (hate speech triage).
- Vision: Eye (image classification), Moderator (content moderation), Shapes (shape recognition), Face (face matching).
- Free tier: Each model is free up to 100k monthly active devices.
Use Cases of Desert Ant Labs
- Speech enhancement: Improve audio quality in real-time without cloud processing.
- PII redaction: Automatically filter sensitive information on-device.
- Speech recognition: Transcribe audio quickly (e.g., 10 minutes in 2 seconds on an iPhone).
- Clip selection: Create video shorts and highlights.
- Image classification: Tag and rank images based on aesthetics and content.
- Emoji suggestions: Suggest emojis faster than typing.
- Shape recognition: Convert rough sketches into perfect shapes.
- Content moderation: Flag inappropriate content before upload.
- Language identification: Detect language from just a few words.
- Filler-word detection: Find and remove filler words from transcripts.
- Speaker labeling: Identify who said what in audio and video.
- Face matching: Find people in photos privately on the device.
- Structured extraction: Extract typed JSON from any text.
- Hate speech triage: Catch hate speech before it posts.
Pricing
Desert Ant Labs offers a free tier: each model is free up to 100k monthly active devices, with unlimited inference. No token costs or logins required.
FAQ
Q: What platforms are supported?
A: SDKs are available for Swift, Kotlin, and JavaScript.
Q: Is there a cost for inference?
A: No, inference is unlimited and free up to 100k monthly active devices per model.
Q: Can I use these models commercially?
A: Yes, the models are free to use within the specified limits. For licensing details, refer to the license page.
Q: How do I get started?
A: Visit the documentation for guides and examples, or check out the SDK on GitHub.
Q: Are the models privacy-preserving?
A: Yes, all processing happens on-device, so data never leaves the device.
