AI Super Agent – One Brief, Every Medium

Made with 7ART AI Super Agent

How 7ART compares

Frequently asked questions

• Can the agent spend my credits without asking?

  No, and the design is deliberate about it. The agent writes a plan; it does not generate. The plan card shows every step and the total credit cost, and the studio only starts work when you press Create. The estimate is computed with the same pricing functions the generation routes deduct with, so what you approve is what you pay — the server stays the final authority on the exact number.

• Which AI models can direct it?

  You choose. The list is Gemini 3.5 Flash (the default) and Gemini 3.1 Pro, Claude Opus 4.8, Sonnet 5, Haiku 4.5 and Fable 5, GPT-5.6 Sol and GPT-5.6 Luna, Grok 4.5, DeepSeek V4 Pro, Kimi K3 and Qwen3.7 Max, plus a set of older ids kept valid so a saved pick never dead-ends. Anything unrecognised quietly falls back to the default rather than failing. The director model plans; the generation models named in the plan are what actually render.

• What can it make?

  Steps can be image, video, music, voice (text-to-speech), 3D, lipsync, motion control, sound effects, foley, video edit / restyle, and translation of a talking video into another language. On top of that it has two long-form jobs of its own: a full manga or comic book, and a website or landing page with generated imagery that downloads as HTML or as a project folder.

• Can it do anything the normal workspaces cannot?

  Yes, a handful of engines are only reachable through it. Upscale, relight and FLUX.1 Kontext instruct-edit are not in the Image workspace picker, and foley and video restyle are not in the Video workspace — they exist only on the agent path. If you have been looking for an upscaler in the Image tool and could not find one, this is where it lives.

• What does it cost to talk to it?

  Conversation is billed on what the language model actually used, settled after the turn from the provider's reported cost, which for ordinary back-and-forth is a small number of credits. The expensive part is always the generations, and those are quoted on the plan card before you approve them. Picking a heavier director model makes the conversation cost more; it does not change what the renders cost.

• Can I use what it produces commercially?

  Yes. 7ART's terms permit personal and commercial use of generated content, subject to those terms, applicable law and any restrictions from the underlying model providers, and the legal status of AI output varies by country. Note that downloading files requires a paid plan — a free account can generate and preview but not export.

• What happens if a step fails?

  The generation is refunded automatically, and the failed step keeps its own Retry so you never re-run a whole plan to recover one shot. The retry is careful with money: if a submit's outcome was never confirmed it re-attaches to the render already paid for, if a render only timed out it re-polls the same job instead of buying a new one, and if an attempt is known to have been charged the card says so and tells you to check your library first.

• How long does a plan take to run?

  It depends on the medium. The estimates the app shows while it works are roughly 25 seconds for a voice line, 40 for an image, 45 for a sound effect, about 100 for a music track, three minutes for a video and five for a 3D model. Independent steps run at the same time, so a plan is usually as slow as its longest branch rather than the sum of its parts. A manga book or a website is many generations in a row, so budget longer for those.

• Do I need to know how to write prompts?

  No. You describe the outcome; the agent writes the prompt — subject, environment, lighting, lens, camera move — picks the engine and says in one phrase why. It asks at most one clarifying question, checks with you before any music or sound effect, and describes a multi-step approach in prose before it emits a plan. If you do know exactly what you want, name the model and it switches to that one.

• Can I attach my own photos, clips and audio?

  Yes, up to six files per message — images, video, audio, PDF or text — uploaded or picked from your 7ART library. Attached images are shown to the director visually. It routes the rest by job: an attached video is what a foley, restyle, dubbing or motion step uses, and attached audio is what a lipsync speaks. If you attach a clip, it plans on an engine that can actually read it rather than one that would quietly drop it and bill you anyway.

• Is there a watermark on what it makes?

  On a free account, images come back with a small "Generated on 7ART.ai" watermark; video, music and 3D do not carry one. On a paid plan there is no watermark, and every result has a download: a step's own file, a manga as a PDF, and a website as a single HTML file or a project .zip.

• Can I change a website after it is built?

  Yes. With the site open, tell the agent what to change and it re-emits only the pages it changed, keeping the images that already exist rather than regenerating and re-charging them. You can also regenerate one asset on its own, replace it with your own upload, or point it at a file from your library for free. Then download the HTML for a single page, or the whole project as a .zip with an assets folder.