Skip to content
AI EmployeeSuper-AgentsAgent-to-AgentBenchmarksPricingBlog

CellCog's 3D Tool Adds a Precise Level

At a glanceQuick answers
What changed?
CellCog’s 3D tool gained a second level, precise, for finer textured models from your photos. Fast is still the default.
How do I use it?
Ask your agent for the precise level and give it one to four photos of the object. It tells you the model takes about 10 minutes before it starts.
Does it cost more?
Yes. Precise uses about 1.7 times the credits of fast.
Editorial data illustration on a near-white ground: a photo of a ceramic teapot on the left feeds two paths, a short teal path labelled fast ending in a smooth low-detail model, and a longer amber path labelled precise ending in a finely detailed textured model, beside the large text 4 views and about 10 minutes.
Fig 0Same photo, two levels of detail. Made by CellCog's image agent, running GPT Image 2.6.

CellCog’s 3D tool now has a second level. Fast stays the default and still turns a text prompt or a photo into a model in 1 to 3 minutes. Precise is new: it builds a finer, always-textured model from up to four photos of an object, takes about 10 minutes, and uses about 1.7 times the credits of fast. The precise level went live on November 16, 2026.

On this page · 6 sectionsOpen
  1. Two levels, side by side
  2. How to ask for it
  3. Where precise is not offered
  4. Opening the file
  5. What we are not claiming
  6. Sources
Key points6 · 4 min full read
  1. Two switches, one flipped up in amber: a second level.
    Since November 16, 2026, CellCog’s 3D tool has two levels. Fast stays the default; precise is new and builds a finer model.
  2. A speech bubble and a photo card feeding a simple cube, with a stopwatch: fast.
    Fast works from a text prompt or an image, takes up to 8 views and returns a model in 1 to 3 minutes.
  3. Four photo cards around an object: up to four views.
    Precise works from images only, takes up to 4 views (front, back, left and right) and takes about 10 minutes. Ask in words and the agent makes an image first.
  4. A fine mesh sphere half painted with texture: textured models.
    Precise always returns a textured model, from 100,000 to 2,000,000 faces. Geometry-only models stay on fast, which goes from 40,000 to 1,500,000 faces.
  5. A file icon beside a preview picture: GLB plus a preview.
    Precise uses about 1.7 times the credits of fast. Both return a GLB file with a preview image.
  6. A shield with a padlock beside a photo card with a stop sign: not offered in some regions.
    Accounts in the EU, the EEA and the United Kingdom get the fast level only. A precise request from one of them makes nothing and charges nothing.

§ 01Two levels, side by side

Fast (default) Precise
Input Text prompt or image Images only; from words, the agent makes an image first
Views Up to 8 Up to 4 (front, back, left, right)
Faces 40,000 to 1,500,000 (default 500,000) 100,000 to 2,000,000 (default 500,000)
Textured or geometry only Both Textured only
PBR materials Optional Optional
Time 1 to 3 minutes About 10 minutes
Credits Base About 1.7 times fast
Output GLB plus a preview image GLB plus a preview image
Table 1Fast and precise, as shipped on November 16, 2026

The two levels answer different jobs. Fast is for drafts, ideas and anything that starts as a sentence: a prop for a scene, a quick look at a shape, a model you will repaint. Precise is for the object you already have in front of you, a product, a part or a sculpture, when the model has to look like the photo.

Animation: photo cards of a vase feed two isometric tracks; the short teal Fast track ends in a plain faceted grey vase, the longer amber Precise track ends in a finely faceted patterned vase
Fig 1Same photos, two levels: a quick simple model, or a slower textured one

§ 02How to ask for it

Tell your agent which level you want. For precise, give it one to four photos: a front view is enough, and back, left and right views help it fill in what the front hides. The agent tells you the model takes about 10 minutes before it starts, so you can keep working while it builds.

If you ask for precise in words alone, the agent makes an image of the object first and builds from that image. If precise cannot take a request, for example an image format it does not read, the agent builds the model on fast, bills it as fast, and tells you so in its answer.

§ 03Where precise is not offered

The precise level sends your images to a provider that may use them, and the model it returns, to improve and train its own AI models, de-identifying them where it can. That does not happen for accounts in the EU, the EEA, the United Kingdom and nearby territories, or for accounts whose country we cannot confirm: those accounts get the fast level, and a precise request from one of them makes nothing and charges nothing. If you do not want your images used this way, ask your agent for the fast level. The Privacy Policy carries the full wording.

§ 04Opening the file

Both levels return a GLB file: the binary form of glTF 2.0, the open 3D format published by the Khronos Group, which keeps the mesh, its textures and any PBR materials in one file. Blender imports glTF 2.0 directly, and Google’s open-source model-viewer shows a GLB on a web page with a few lines of HTML. The preview image that comes with each model is there for a quick look before you open it anywhere.

§ 05What we are not claiming

We do not compare precise with other 3D tools or publish benchmark scores. We do not promise print-ready or game-engine-ready output, or any format other than GLB. Times are approximate: the first production model on precise took about 9 minutes.

§ 06Sources

Frequently asked5 questions

Q1Can the precise level make a 3D model from text?

Not directly. Precise works from images only, so when you describe an object in words your agent first makes an image of it, then builds the model from that image.

Q2How long does the precise level take?

About 10 minutes, against 1 to 3 minutes for fast. Your agent says so before it starts, and you can keep working in the meantime.

Q3Can precise make a model without textures?

No. Precise always returns a textured model. If you want geometry only, for example to paint or print it yourself, ask for the fast level.

Q4Why is precise not available in the EU, the EEA or the UK?

The precise level sends your images to a provider that may use them, and the model it returns, to improve and train its own AI models. We do not send images from accounts in the EU, the EEA or the United Kingdom, so those accounts use fast. Our Privacy Policy has the details.

Q5What file do I get?

A GLB file and a preview image, on both levels. Optional PBR materials can be added on either level.

Published 16 November 2026 All Changelog →