Tutorial, use cases and news
A model's API page is now a real reference for what it supports, laid out right in the UI. No more fetching anything yourself just to find out.
Modern TTS goes beyond basic speech synthesis: preset voices, voice cloning, and voice design, each with different quality, control, and speed trade-offs.
How to choose an embedding model for RAG: weigh retrieval quality, dimensions, latency, and language support, not just benchmark scores.
We’ve learned what makes people thrive at WriftAI, from action bias and ownership to intellectual humility and teamwork.
Cancel a running WriftAI prediction from the API or any SDK. Useful for stuck predictions and slow models like video generation, with no queue-time charge.
A practical guide to prompt conditioning for diffusion models: structuring prompts for predictable, controllable image outputs.
WriftAI now supports enterprise Single Sign-On (SSO) with SAML 2.0 and OIDC. Learn how SSO works and how to set it up for your organization's domain.
See the status, URL, and time of every webhook we sent you, right in your dashboard. Filter down to just the failures when something is not arriving.
The WriftAI CLI now speaks OAuth 2.0 device authorization. Log in from your terminal, approve in your browser, done.
Between launches, deadlines, and product milestones, we make time for the experiences that bring our team together.
WriftAI rate limits are now documented. This post covers the limits by tier, the response headers, and how to implement correct 429 retry logic.
WriftAI is switching from end-of-month invoicing to prepaid credits on July 1. Here's why, how the new system works, and what to do before the switch.
WriftAI model pages now have a playground and version history. Run any model in the browser before writing code. See what changed between versions.
POST /v1/predictions now accepts a webhook field. Pass a URL and WriftAI will POST to it the moment your prediction completes. No polling needed.
WriftAI's API now has a public OpenAPI 3.1 spec. Import it into Postman, generate clients, and run mock servers.
POST /v1/predictions now returns immediately. The prediction runs in the background. Here is the lifecycle, how to poll, and why this was necessary.
L40S is now available on WriftAI. The right fit for many image generation, audio transcription, code completion, and document models.
A problem we kept running into at Sych: every ML team spending weeks on infrastructure before a single model could run. That's why we built WriftAI.
Scale your projects effortlessly with WriftAI. Seamlessly integrate and optimize performance as you expand and innovate.