Documentation Index
Fetch the complete documentation index at: https://docs.opencanva.co/llms.txt
Use this file to discover all available pages before exploring further.
New features
Video-aware prompt enhancement — The prompt assistant now optimizes your prompts differently for each video model. Kling prompts gain cinematic vocabulary and camera motion cues, Sora prompts use natural-language enrichment without injecting camera instructions, MiniMax prompts skip enhancement when the model’s own optimizer is active, and Veo prompts are structured with subject, context, action, and style fields. Other video models get a balanced default. This happens automatically when you enhance a prompt with a video model selected.Updates
Multi-provider generation routing — Image generation now routes requests across both fal and Replicate providers. The system picks the right provider for each model automatically, so models like Runway Gen-4 Image that run on Replicate complete inline instead of waiting for a webhook. Cancelling a generation also routes to the correct provider. You don’t need to change anything — the right path is selected for you.Stronger video upscaling pipeline — Topaz video upscaling now handles more source formats, supports uploads alongside generated videos, and detects resolution more reliably. Mux-hosted videos prefer the MP4 rendition for upscaling, and pricing thresholds are consistent across quality levels.Bug fixes
- Prompt enhancement validation — Fixed an issue where video duration and quality settings could be missing from the enhancement context, leading to generic results instead of model-specific ones.
- Generation cancellation — Cancelling a generation now correctly cleans up on the right provider, preventing orphaned jobs that could consume credits.