Nothing yet—I'm severely disappointed with the experience of using it. Review collected by and hosted on G2.com.
My biggest issue is the lack of transparency around OpenArt Director’s credit consumption. I ran a very simple 6-second, single-shot vertical video using an existing character. Director explicitly said it skipped the full script/planning workflow and went straight to a one-shot render. Despite that, the generation still consumed 1,304 credits in total.
In my credit history, 652 credits were for the actual media generation (GPT Image 2 + Seedance 2.0), and another 652 credits were categorized as “Director - Agent Conversation.” In other words, 50% of the total cost went to agent/conversation overhead, even though I barely interacted with the agent and had chosen the most direct generation path.
I find this billing model extremely hard to justify. The interface didn’t make it clear upfront that a short, simple Director request could trigger hundreds of additional credits in agent charges. Users should see an estimated breakdown of generation cost versus Director/agent overhead before committing to a render.
I also ran into a safety-filter issue: my non-explicit character reference sheet, which included ordinary swimwear reference images, was rejected. That forced Director to fall back to a different reference image, which reduces the usefulness of the character-consistency workflow for fashion, swimwear, and creator-style content.
OpenArt has strong underlying models and useful tools, but Director’s AI agent - hidden orchestration costs seriously hurt ROI. For now, I’d recommend using the direct image/video tools instead of Director unless OpenArt makes these charges transparent and substantially reduces the agent overhead. Review collected by and hosted on G2.com.
