Lovart¶

Development line: project:lovart · thread lovart-development
Last event: - · 0 dated since - · Researched: 2026-09-05 · confidence: medium

What it is¶

Lovart is a web workspace for generating brand images, video, and graphics through an agent or direct model selection.

• Canvas: arranges and edits assets on an open workspace.
• Brand Kit: manages logos, colors, fonts, and guidelines, or parses brand-book PDFs.
• Point editing: adjusts specific elements directly.
• Image and video generation: generates media via an agent or direct model selection.
• SVG vectorization: converts PNG and JPG raster images into SVG.
• PSD export: exports canvas contents to layered PSD files.

Canvas generates SVG files but cannot edit vector paths; path editing requires Illustrator, Figma, or another vector editor. It serves as a working layer over multiple generative models rather than a replacement for a vector path editor.

Development line¶

• The dated line is not written up yet; what is known stands in the sections below.

What changed¶

• 2025-05-13: Lovart ran in closed beta; a later primary post confirms the beta started in May 2025.
• 2025-06-05: added Flux Kontext integration to edit images by text and references.
• 2025-06-11: added Veo 3 for video generation.
• 2025-06-18: added Kling 2.1.
• 2025-07-23: added Chat with Canvas to use comments on an image or area as agent input.
• 2025-07-28: exited beta for public release alongside ChatCanvas; the company reported over 300 000 early users.
• 2026-02-19: added Vectorize to convert PNG and JPG into SVG.
• 2026-03-05: added cropping presets for social platforms.
• 2026-03-09: found no verified primary information on a specific release in accessible links; short link hf.ru was not verifiable as a Lovart source.
• 2026-03-21: added toolbar customization.
• 2026-03-23: added moving selected objects inside an image with generative fill for the vacated area.
• 2026-03-27: found no verified standalone delivery for this date; nearest confirmed changes are Move Object on 23 March and PSD export on 29 March.
• 2026-03-29: added export of Canvas contents to layered PSD.
• 2026-09-05 status: Lovart displays integrations for GPT Image 2, Seedance 2.0, Nano Banana Pro, Kling 3.0, Seedream 5, Veo 3.1, Midjourney, Luma Uni-1, and Flux 2 Pro; availability and settings depend on the selected model and plan.

How to use this¶

No practitioner action is proposed: on 2025-05-13, 2026-03-09, and 2026-03-27, the dated links alone do not establish a verified Lovart product-development change.

1. Create a new project or start Quickstart; describe the task in plain text and check the result on Canvas. — https://www.lovart.ai/docs/getting-started/design-your-first-project
2. Set up Brand Kit first for an existing brand: upload logos, colors, fonts, and guidelines, or let the service parse a brand-book PDF; apply the kit to the project. — https://www.lovart.ai/docs/getting-started/things-you-should-set-up-first
3. Open Models for guided generation: Auto selects a model, Auto Off sets preferences, and an @-mention fixes a specific model for a single prompt. — https://www.lovart.ai/docs/how-to-prompt/selecting-ai-models
4. For SVG, select a PNG or JPG on Canvas, choose Vectorize, download the SVG, and open it in an external vector editor if needed. — https://www.lovart.ai/docs/edit-your-design/ai-transformation
5. Export a single object or select multiple objects; batch export supports PSD. — https://www.lovart.ai/docs/getting-started/design-your-first-project

Best practices¶

• Build the Brand Kit before first generations for an existing brand, and maintain a separate kit for each brand. — https://www.lovart.ai/docs/getting-started/things-you-should-set-up-first
• Treat model preferences in auto-selection as preferences rather than strict locks; fix a specific model with @ when needed. — https://www.lovart.ai/docs/how-to-prompt/selecting-ai-models
• Remove backgrounds before Vectorize so logos and icons trace cleanly. — https://www.lovart.ai/docs/edit-your-design/ai-transformation
• Outline objects tightly for Move Object edits: wide selections hurt recognition in complex scenes. — https://www.lovart.ai/docs/edit-your-design/ai-transformation

Superseded by this¶

• 2025-05-13: guidance on closed-beta access was superseded by the public launch on 2025-07-28.
• Before 2026-02-19: the assumption that Lovart does not convert raster to SVG was superseded by Vectorize.
• 2026-02-19: the assumption that Vectorize replaces a vector editor is unsupported; Canvas does not yet edit SVG paths.

Still unknown¶

• Dates 2026-03-09 and 2026-03-27 lack verified descriptions of specific releases in accessible links; we cannot link them reliably to Vectorize or neighboring changes.
• Short link https://hf.ru/linkd4c82 remains unused: we could not verify its target material or connection to Lovart.
• Lovart marketing pages state capabilities and user counts; no independent audit has verified quality, credit limits, or model availability.

Sources¶

Agent brief¶

• Subject: project:lovart, thread lovart-development, 0 dated events - → -.
• Practical note: No practitioner action is proposed: on 2025-05-13, 2026-03-09, and 2026-03-27, the dated links alone do not establish a verified Lovart product-development change.
• Confidence: medium. Dated supersedes above are the authority for what is obsolete.