More 3ds Max files convert
Newer .max files now open. MeshObject and Body Object geometry is read, and shapes become tubes, bars, flat faces or extruded solids as 3ds Max exports them. Parts keep their correct rotation.
Newer .max files now open. MeshObject and Body Object geometry is read, and shapes become tubes, bars, flat faces or extruded solids as 3ds Max exports them. Parts keep their correct rotation.
Pages now stay up when Chrome's page translation is on, across the whole site. In rare cases an old translated label can stay on screen a little longer than it should.
SolidWorks faces of any size are now read, so large curved faces and imported meshes no longer vanish.
Sketch-only parts now say they have no solid body.
Roblox exports now split meshes into smaller pieces, upload each distinct mesh once with retries, and place parts in world space. RBXM and RBXL exports finish instead of timing out.
Imports accept .rbxmx and .rbxlx and read large places much faster. They keep animation poses and read wedge, corner wedge, truss and seat parts.
SolidWorks files saved by releases up to about 2008 now convert. Convert3D decodes the older compression method these files use.
Your model stays on screen even when the lighting map can't be loaded, for example on networks that can't reach our asset host. A built-in fallback lights the model with no download.
This covers the converter, viewer, render page, compressor and Image to 3D.
SketchUp conversions now match SketchUp 2026's own importer and exporter more closely. Sizes, materials and soft edges line up, including USD and USDZ models in centimeters, glTF models with all their materials, and 3DS models.
JPEG textures stay JPEG, and an image shared by several materials is stored once, so files can be much smaller. OBJ and USDZ exports from SketchUp are cleaner, with fixes for glass materials, mirrored parts and names.
On a phone, Image to 3D stacked the composer, the viewer and the history on one page.
It now works like the converter, one screen at a time. First the composer, then a full-screen live preview with a progress bar, then the finished model with Done, the file name and Download. A back arrow returns to the composer while a run continues. Renaming and rating are desktop-only for now.
Files up to 256 MB are now kept in memory when you add them, so you can convert them again even if the original, such as a temporary unzipped file or an email attachment, is gone.
Blend1 and bare .model files import correctly. Large 3MF exports finish, including from very large STL files.
STEP and IGES files now read about twice as fast, and the downloaded engine is about 7 MB smaller.
Results are also steady: the same file always gives the same model.
SLDDRW files now convert to 3D. Convert3D reads the 3D model that the drawing's views are shaded with, since a drawing's own display data holds only the 2D sheet.
Part drawings become one mesh per part, with sheet-metal parts kept in their formed shape. Assembly views are read too.
SketchUp conversions that produce a zip of several files now finish.
A SketchUp model with a very large photo as a texture converts too. Convert3D skips that one texture and converts the rest of the model.
IFC building models with a very large number of parts in one color now convert.
Large 3DM, 3MF and STEP files now convert in seconds.
Some 3DM files that failed before now convert. FBX results are smaller, and large SketchUp files use less memory.
Image to 3D used to come with usage limits. It is now free and unlimited for everyone.
Files over 2 GB now import in Chrome. Big files are read in slices, and STL files are read in chunks, so file size is no longer the limit. OBJ and G-code read in slices as well.
STEP output is also welded and sized more tightly.
DAE files now export with their geometry, materials, transparency, normal maps and textures. Results look the same in Blender at a smaller file size.
DAE import reads more files, including skinned models with joints outside the skeleton and COLLADA 1.5 textures.
In the ChatGPT app, you can reopen an attachment after its link expires, and the "Convert on convert3d.org" button works, because a saved copy is now kept.
The app also runs in the ChatGPT desktop app.
You couldn't convert Revit .rfa or .rvt files before. Now you can import them and convert them to other 3D formats.
Solids are read straight from the file, and a family placed many times is stored once. Results closely match Revit's own OBJ export.
Image to 3D now opens on iPhone and keeps your image. Add an image on the landing page, open the workspace, and your image is there until it expires.
The larger model is not offered on phones and tablets, which use the lighter one. Before, iOS could shut the tab down with "Your selected images could not be restored".
The lighter Image to 3D model now takes about half as long. It needs much less to download and far less memory while loading.
It runs on any WebGPU device, including iPhones, and shows a live preview while it works.
Models from STL, 3MF, AMF, G-code, STEP, IGES and BREP files now convert at their real size and stand upright in GLB, SKP, FBX, USDZ and DAE.
STL to STL and other formats without units keep their numbers exactly.
Slicers read STL and 3MF as millimeters with Z up. Models from GLB, glTF, SKP, FBX and MAX are now written that way, so a 1 m model arrives 1000 mm tall and standing upright.
A tool that reads STL as meters, such as Blender's STL import, will now show them 1000 times larger.
Each converted SKP now gets its own model ID, so you can import several converted files into one SketchUp model.
On Windows, STEP, IGES, BREP and DWG files now convert in the SketchUp converter too.
Rigged .blend characters now keep their saved pose and size, and parts attached to bones hang from the right spot. A mirrored half is bound to the opposite side's bones, so both sides line up.
Fixed an issue where FBX files could come out too big for a small subset of users.
Pop-up messages across the site now appear. That includes card payment failures, failed Paddle checkouts, "Generation cancelled" and "Could not remove model", and messages sent while a page is still loading.
Large files now upload from the ChatGPT app on a phone. If an upload fails, the app sends the whole file in one request.
If converting, uploading or opening an attachment still fails, the app shows a button that opens the file on convert3d.org, which converts it right away. Errors say which step failed.
A 3MF file can say its numbers are in meters, inches or another unit. Convert3D now applies that unit, so models keep their real size.
Files in millimeters, and files that don't state a unit, are unchanged.
STL files are now binary, at 50 bytes per triangle instead of about 250. Files are about a fifth of the size, and very large models convert.
Converting G-code to STL, 3MF or OBJ now builds the printed part as a solid. Convert3D reads the extruded moves and uses the line width and layer height from your slicer.
G-code with no extrusion, such as CNC, still exports its toolpath. Files ending in .gco, Cura's default, are accepted.
Cut-off IFC files, such as a Revit export that stopped partway, now convert as far as they are readable. An empty file gets the message "The file ... is empty."
One bad IFC file no longer affects the next one you convert in the same tab. The API reads these files the same way.
G-code from CNC, laser and plotter programs now exports its toolpaths as lines to DXF, DWG and OBJ. Arcs, inches and compact coordinates like "G0X0Y0" are read correctly, so drawings keep their size.
G-code now stays Z up, so a print stands upright in STL and 3MF.
The ChatGPT app now zips the converted file before handing it over, which ChatGPT's file storage accepts. The usual upload stays as a fallback.
Tapping Download in the ChatGPT app now works even on networks that can't reach the storage host. The file goes up in parts, with retries on each part and a wait for the page to be visible and online.
If parts keep failing, the app sends the whole file in one request. If something still goes wrong, the error says which step failed.
DWG and DXF files from STEP and SolidWorks parts now carry real 3D solids with exact surfaces. AutoCAD 2027 opens them, accepts them as valid solids and measures the same volumes.
STEP export now writes a solid too, with each shared edge written once, so other CAD tools open it as a solid instead of loose faces. Before, DWG and DXF exports arrived as wireframe.
When a large STEP file runs out of memory midway through meshing, the conversion now fails with "OCCT ran out of memory meshing this file, so part of it is missing." A paid job moves to a larger server and finishes the full model. Before, the conversion could report success with holes in the model.
Image to 3D model downloads now retry and resume on a flaky connection. Devices with no WebGPU are told so up front, before the 17.5 GB download starts. GPU problems now report what actually failed, including the GPU timeout some Windows machines hit.
In the ChatGPT app on an iPhone, the viewer toolbar now shows only home, fit, pan, zoom and orbit, at the top. Progress reads "Converting to OBJ... 50%" instead of internal step names. The download upload retries up to three times if the connection drops.
The API server now watches its memory and stops the heaviest conversion before it uses up everything, so other conversions keep running. A paid job moves to a larger server. Other jobs fail with the message "This file needs more memory than this server has."
API conversions can now run for up to 29 minutes, up from about 290 seconds. Large files that need a few more minutes finish.
Large STEP files now import with far less memory, and a hang on some files is fixed. Files that still don't fit have a fallback with a larger memory limit in browsers that support it.
A paid conversion that needs more memory than the API server has now moves to a larger server, and its progress reads "Moving to a larger server". This lets big STEP files finish. Free conversions are not moved. If the move is not possible, the conversion fails as before.
Very large models now export to DWG, including files of several hundred megabytes, which write in full. If the server really runs out of memory, the message says so.
Files ending in .blend1, which Blender saves as backups, now convert as Blender files. Files ending in .model, which sit inside an unzipped 3MF, convert as 3MF. If a .model file is only a reference to other parts, or is not 3MF at all, you get a clear message. Blender 5 files now convert with their contents instead of an empty GLB.
A .3ds file refers to its texture images by name, often in capital letters or with a folder path. Texture files are now matched by name without regard to case or folder, so a texture you upload next to the model is found. A texture you did not upload is left out without causing a problem.
Convert3D now reads .3ds files with a more forgiving reader, so models with a small defect, such as extra data at the end of the file, convert. Files with no geometry at all get a clear message.
The 3D viewer and render page now load their lighting presets from Convert3D itself, so models are lit on more networks. They used to download the presets from an outside host.
Landing pages now work when your browser translates them, for example with Google Translate. Drop a file and the conversion starts as usual. Some visitors used to see an "Application error" at that point.
The lighter Image to 3D model now runs on phones, older GPUs and Windows machines with limited graphics memory. Background removal falls back to the processor when the GPU can't do it, and the final step fits into small memory blocks. Devices with under 4 GB of RAM are still turned away.
A .gltf file often depends on a separate .bin file. If you upload the .gltf alone, the conversion now says which .bin file is missing. It tells you to upload it together with the .gltf, or to zip them and upload the ZIP. The failure dialog shows this message for every format.
Large DWG drawings with many repeated blocks, such as furniture symbols or fixtures, now convert faster and come out smaller. Each block is drawn once and placed wherever it is used.
The API now converts 3DM and IFC files. DWG and DXF text also uses a font that ships with Convert3D, so conversions no longer fetch one from another site.
FBX objects that share a shape now keep their own materials in each place. If one copy is red and another is blue, each stays that colour. The shape is still stored once, so the file does not grow.
An FBX file that places the same object many times, such as trees or fixtures in a building, now converts to a much smaller GLB. The object is stored once and placed each time. The result has the same models in the same positions.
Scenes with many copies of the same object, such as trees, fixtures or repeated building parts, now export to FBX at a sensible size. Shared geometry and materials are written once. Some files that used to run out of memory now export.
Large STEP and IGES files now keep running until they finish. The import reports progress as it reads, builds and meshes the model, and a few slow steps are faster too. The converted files are the same as before.
Models with KTX2 textures now keep them. The textures come through as regular images, in the browser and through the API, so the converted model keeps its look. Before, the textures were dropped without a warning.
FBX files made from a GLB now keep rotated nodes, rigs and objects held by a bone where the GLB had them. This helps characters from Blender and Mixamo. Animation follows.
Large SketchUp files now import and finish, including older files from version 8 to 2020. The slowest steps of the import are much faster, and the longest quiet stretch is a few seconds. The converted result is unchanged.
Very large models now export to FBX, STEP, OBJ and GLB without being stopped. These exports keep reporting progress while they write, so long exports run to the end. The files you get are the same as before.
The API now reads Draco-compressed GLB files and converts them to any format. This matches the browser, which already handled them.
DWG files now convert to GLBs that are several times smaller. They build in seconds, and progress keeps moving on the API. Some conversions used to time out while building the GLB.
Large STL files now convert to GLB. Import uses far less memory and is much faster.
IFC exports now contain only the model you converted. Before, a conversion in the same browser tab could include IFC files you had converted earlier, so a box converted after another came out as both boxes.
Rigged FBX characters now import with their proportions intact. Convert3D reads the skeleton correctly, so they match Blender's own import. Some rigged characters used to come in warped.
FBX to GLB conversions now keep their textures. Materials with no emission stay unlit, so textures show instead of a white glow.
Each 3DM conversion now keeps its own materials, textures, material lists and warnings. Some conversions used to show another file's material and texture on your model.
STEP files with an edge that has no curve data now convert to DWG and DXF. Convert3D skips that edge and converts the rest.
SLDPRT files built from imported meshes now convert to STEP. These parts go to STEP as their mesh. Some of them used to stop with an error, even though their GLB worked.
Roblox models with a part inside another part now keep every piece where it belongs. Nested parts sit at their stored position, so a character comes out whole in the viewer.
An OBJ file now converts even if it names a texture you did not upload, or one whose image cannot be read. Convert3D leaves out that one map and converts the model with the rest. Before, one bad map could stop the whole scene.
3ds Max files with textures now convert to GLB. Convert3D waits for each texture image to load before it exports, so every texture is ready when the model is written. Some files used to stop with an error.
Large COLLADA files now convert to GLB and 3DS. They finish without timing out.
Convert3D copies the file in one step instead of one byte at a time, which was what slowed loading.
FBX files from Convert3D now have upright textures, open at their true size in Blender, and keep their colors on a round trip. FBX files from other tools may import slightly darker, which is correct.
Large models export to DWG much faster. Each line takes about the same time to add, so big models finish well within the time limit.
GLB files made from other formats are now clean. Leftover memory bytes are zeroed, so the same input always gives the same file, and API results never include bytes from another conversion. The model itself is unchanged.
3MF files with materials or colors now convert, and so do AMF files. Colors come through and all objects in a file are kept.
3DM files with embedded textures now convert to GLB. Convert3D waits for the textures to load before it exports.
3DM files whose reflection environment is a single color now load as usual.
GLB models that pack their vertex data together or in compact form now export to USDZ correctly, with the right normals, UVs and vertex count.
Textures are now read in a way that every exporter can handle. Exports that include textures finish, and textures come out the right way up.
Conversions and previews work in older browsers, including Chrome 109 on Windows 7 or 8, Firefox 115 ESR and Safari 16.
Large models, such as big OBJ files, now export to USDZ.
Large conversions to OBJ now finish. 3DM exports work, and DXF files convert. If a STEP or IGES file can't be read, you get a real reason. DWG and DXF to DXF now go through a reader checked in AutoCAD.
You couldn't open or convert a 3D file in a ChatGPT chat before.
Convert3D now runs as a ChatGPT app. Attach a model to the chat, see it in a viewer and convert it to one of 15 formats. SKP, DWG, STEP, FBX, USDZ, Blender, DAE and SolidWorks files are among those it reads. It uses the same conversions as the website, and a button lets you download the result.
The viewer loads the right model after you leave Compress, whether you press Back or switch apps.
The Convert3D viewer in ChatGPT now uses the same drop area and preview as the website, so zips and models with textures work, and GLB files show up. One button goes from Convert to Converting to Download, and the download opens from your click.
If a file can't open, you land back in the drop area.
Large conversions now download in ChatGPT. Downloads go through a secure link, so big results download without trouble.
If a computer can't run the high-fidelity Image to 3D model, you now see "This device couldn't run the high-fidelity model" with a button to use a lighter model. It runs the same image again. Results are less detailed.
Choosing an example on the Image to 3D page now skips the intro card and starts once the model files are ready. "Not now" closes a card without downloading, and Cancel closes the card during a download.
SolidWorks assemblies now convert with every part in place, in GLB and STEP. Each instance sits where the assembly puts it, so repeated parts such as bolts and mirrored copies appear. Hidden and suppressed parts are skipped.
Parts and assemblies go to STEP as solid shapes when the file has them, and as meshes when it does not.
The Image to 3D progress bar now moves through each stage, so you can see how far along your model is.
Returning users also get a card when a model update is available.
Image to 3D now runs on more computers. A lighter model works on other WebGPU devices, and a note says results are less detailed while it downloads. Background removal also works on GPUs that lack some advanced features.
Image to 3D finishes sooner. Loading is quicker, and the texture takes fewer steps.
The texture looks slightly different. In a blind review it was judged as good.
If your connection drops while Image to 3D downloads its model files, it now reports a network problem and you can try again. The "This browser can't store the whole model" card appears only when your browser really is out of space.
You can now try the MCP endpoint with no API key. Anonymous calls can list formats, convert and check status, up to 25 MB per file and 10 conversions an hour per IP address. They do not use credits.
Uploading through the endpoint still needs a key, which comes with a paid API subscription.
Older SketchUp files (version 8 to 2016) now draw their curved faces correctly. A text record in the file is read properly, so everything after it lines up.
SketchUp files now convert the way SketchUp shows them. Textures keep their direction, transparency reads correctly, colorized textures match SketchUp, and only components the file places are drawn. We checked the results against Trimble's official SDK.
Exported SKP files are far smaller because repeated parts are written once and placed again. TGA and PSD textures are now read too, in SketchUp, OBJ, glTF and MAX files.
AI assistants had no direct way to use Convert3D. You could only reach it through the REST API.
There is now an MCP endpoint at api.convert3d.org/mcp. An agent can list formats, upload a file, convert it and check the result. It signs in with the same API key as the REST API. Conversions also run in a queue now, so one long conversion no longer holds up everyone else on the API.
Image to 3D can now run entirely on your GPU. Nothing is sent to a server, you do not need to sign in, and there are no credits to spend.
Drop an image, press Generate, and a GLB appears in your history with a Download button. The model has color on its surface. It takes about six to eight minutes on a recent Mac. The first run downloads about 2.8 GB of model files, which your browser keeps for next time. Computers without WebGPU are told so instead of failing quietly.
Image to 3D on your device is faster in the default mode, with the same output. The live preview keeps moving while the model builds.
The Boost switch in Image to 3D came with a text label that was easy to miss.
It is now a lightning button beside Generate. Hover to see "1.5x speed" and a note that it may slow other apps. The add-image button also says "Queue additional models".
DWG drawings made of 3D solids now convert with the right shape. Solids are read from the file's own data, including their curves and surfaces, so curved surfaces stay curved. They are placed where the drawing puts them, and the view frames the part instead of empty space.
Detail is judged across the whole drawing, so a small part no longer looks far more detailed than the rest. Solids placed inside blocks also reach STEP. The results are closer to AutoCAD's own output.
Making a model on your own GPU now leaves room for the page and your other apps. The work runs in smaller pieces with short pauses, and it frees memory as each stage finishes.
The preview updates at a steady, lower rate, and the image preview appears sooner.
The compress page had no way to pick the image format for textures.
A new Texture format menu under Advanced offers WebP, JPEG and PNG. WebP is smallest and keeps transparency. JPEG is the most compatible but has no transparency. PNG is lossless and largest. The model compresses again when you change it. Only color textures use your choice.
The compress page loads your preview once and then compresses. After "Reading file (100%)" you go straight to your model. Some files used to start over at that point.
Models waiting in the Image to 3D queue now show your source image with a Queued label. The model that is running stays in view when you add more images.
The empty history pane and the Text to 3D menu entries are gone.
You can drop up to dozens of images into Image to 3D at once. The button reads "Generate N models", and each image joins the queue and is turned into a model in turn.
If some files in the drop are not valid images, the good ones are kept and the rest are flagged.
FBX export now writes every texture map, linked to the right material, and keeps skeletal and morph animation.
If the new writer can't handle a file, Convert3D still produces the FBX the old way. The update also fixes the direction a few imported FBX files faced.
FBX files keep their true size when you import them, so you can convert a model back and forth between formats and it stays the same size. A 2 m cube stays 2 m.
A recent FBX update had briefly shrunk models on import. That is fixed.
Large 3ds Max and Blender scenes now export to FBX and other formats.
The same update brings back support for RAW, TER, IRR, IRRMESH, M3D and X3D files.
The compress page now starts as soon as your file is read, with no Compress button, and it runs again each time you change a setting. The before-and-after view is the page itself. It also saves more space than before.
A new Large to Small slider picks how the geometry is packed. Large is the most compatible. Small makes the smallest file but needs a viewer that can unpack it.
3MF exports are now compressed, so the files are smaller. Opening one in Convert3D again gives the same size and position as the original. Earlier exports were stored uncompressed, and reopening one could place the model in the wrong spot.
IGES and BREP files now convert to STP with their exact surfaces, curved ones included, so CAD tools can edit the solid. Before, the model was flattened into a mesh of triangles.
Dense 3MF files now convert to every output format. They convert to GLB in seconds, and to STP with all triangles kept. Some used to crash the converter.
Meshes from 3MF, STL, OBJ and similar formats now export to STP directly. The files are much smaller, and CAD tools read them faster.
On phones, the preview now fills the screen below 768 pixels wide, with a back button and one Convert button. Convert opens a bottom sheet where you search for a format. The result shows the file name and a Download button, with options to convert another file or go back to the model. Batch conversion keeps its stacked layout.
Convert3D now checks a binary STL before reading it. If the file is truncated or corrupt, you get a clear, descriptive error. Valid STL files read as before. Some damaged files used to crash the importer.
The format list now shows the last three output formats you picked at the top, above the ranked groups. This saves scrolling when you convert to the same formats often. Recents stay in your browser.
Output formats in the Convert to sidebar are now ordered by popularity, with space between groups. The most common formats get a keyboard shortcut, shown on the card. Shortcuts are off while you type or search.
More SLDASM files now give a model. Convert3D reads the pre-built mesh data that newer assemblies carry. Some assemblies used to fail with no geometry found.
Some dense assemblies can show triangulation glitches. That is still better than no output.
DWG files saved by AutoCAD 2018 that hold 3D solids now convert with their geometry. Convert3D reads those solids. Before, these files produced no geometry.
3ds Max files now convert at the right scale, so they line up with other assets in Blender and elsewhere. Convert3D reads the scene units and converts to meters. Files already in meters are unchanged.
Before, models were written in the units they were saved in, so a file drawn in millimeters came out far too large.
FBX files and CAD assemblies with tens of thousands of objects sharing the same name now convert. The preparation step is fast. Some of these files used to be stopped there.
Some 3ds Max files that use Corona materials now convert with their real colors. Convert3D reads the diffuse color correctly, so materials no longer come out pure red.
Models from 3ds Max scenes that use Corona materials now convert with the right look. Only materials that really emit light glow. Before, every material got a white glow and looked washed out.
Big FBX files now finish converting. Convert3D reports progress during the read, and FBX can use more memory, so slow but healthy files are no longer stopped.
SLDPRT parts with shallow curves keep their shape. The curves are now exact. Some parts used to come out far too big.
SLDPRT files now import as true solids. Convert3D reads the part's actual solid geometry, for files saved before and after SolidWorks 2015, so you keep detail that rough display data lacked. That geometry also feeds STEP export. Display data is still the fallback.
Large STEP, IGES and BREP files now finish converting, even on slower computers. Convert3D reports progress throughout the read, so a valid file is no longer mistaken for a stuck one. CAD formats also get a longer time limit before a conversion is stopped.
More FBX files now convert.
Files that are really another format renamed to .fbx, such as GLB, glTF, OBJ, Roblox models, or a ZIP of a model and its textures, are recognized and converted. Meshes the file did not attach to the scene now appear, and so do point clouds. Revit and Archicad models with brackets in node names convert. Files with circular hierarchies convert when geometry can be recovered. Files that cannot convert, such as XPS or HTML, get a clear error.
The FBX reader was also updated to a newer version.
SKP export now works through the developer API. The API is deploying updates again, so it runs the current version.
Large RBXL asset packs with thousands of mesh parts now finish converting. Convert3D fetches each mesh once, even when many parts use it, and loads several at a time. Some packs used to take too long and fail.
Convert3D now reads Roblox models and places saved as XML as well as in the binary format. Standalone .mesh files import too. Roblox Studio can save either kind of file.
Roblox RBXM and RBXL models import, including models that link their meshes by web address and models with very large meshes. Convert3D now fetches the right mesh for linked models. Some of these files used to fail.
USDZ models that use more than one set of texture coordinates now look as they should. Each texture uses its own set, so textures no longer come out tiled or doubled.
STEP and IGES files now keep their real size in both directions. A file in millimeters imports at the right scale, and a part exported to STP opens at the same size in other CAD tools.
If you converted STEP or IGES files earlier, expect them to come out at their correct scale now.
USDZ files with many large textures now finish converting. Convert3D reports each stage, from loading the USD module to finalizing the GLB, so a slow conversion is no longer mistaken for a stuck one. Some of these files used to be stopped mid-conversion.
Big DWG and DXF files now finish converting, even when a step takes a long time. Convert3D reports progress throughout import and DXF export, so a slow but healthy conversion is no longer mistaken for a stuck one. Some big files used to be stopped part way.
3MF files whose parts are spread across several sections now convert. So do assemblies that refer back to themselves. Some 3MF files from slicers and CAD tools used to come out empty or fail.
When a 3MF cannot convert, the message now says why. A sliced print job from Bambu Studio or OrcaSlicer holds G-code, not a model, so re-export it from the original design. A file that is not a ZIP archive says so, and a file with no geometry says that too.
The viewer now has a toolbar with camera controls, including fit to view and home, plus View and Light menus. Choose a display mode and a lighting preset, or upload your own HDRI environment image. Before, it had one dropdown for material and wireframe, and fixed lighting.
USDZ models with several materials per mesh now convert to GLB in the browser with their textures. Some of these models used to fail, and their textures were skipped.
USDZ files that assign several materials to one mesh now convert, and each part keeps its material.
Files with duplicate or invalid object names, common in Sketchfab and FBX exports, now open. Large scenes export faster. If a USD file is damaged, Convert3D reports the problem and the next file converts normally.
Compressed USDZ and USDC files import and convert. Models that assign different materials to parts of one mesh keep those materials. Some compressed files stopped importing after the USD engine update, and that is fixed.
USDZ files from Convert3D now match files made with Apple's own USDZ tool, so they show up correctly in Quick Look. Models that reuse a material no longer carry duplicate copies of it, and exported files open again reliably. Some earlier files were valid but looked wrong in Quick Look.
This release also updates the USD engine behind USDZ and USDC conversion.
GLB files with a tall or wide texture now convert to USDZ. Some of them used to fail on our servers.
GLB files with textures now convert to USDZ in your browser as well as on the server. Some of them used to stop in the browser while the same file worked on the server.
Transfer is now also a place to review a model. Invite collaborators to a project, leave comments on the model, and mark a comment as resolved when it is dealt with. Before, Transfer only sent a 3D file to someone.
You can now copy preview actions and version tests from another Transfer project you own, so you don't set them up again by hand. Choose "copy from" on a project, pick the source, and select what to bring over. Convert3D asks before it overwrites anything.
Large 3ds Max files now open and convert. Some of them used to stop with an error when Convert3D read their internal structure. It now reads that structure in the right order.
When a conversion fails, Convert3D now shows the usual failure dialog and a notice, so you know and can try again.
In some cases, such as a tab left open across an update, the page had stayed on "Converting" with no error.
You can now zoom reference images in a project with pinch, wheel or the zoom buttons, drag to pan and reset the view. Convert3D remembers the zoom for each attachment.
When a DWG file can't be read, Convert3D now names the cause, such as a missing file section or an invalid DWG file. Several problems in one file are listed together.
Some files had failed with only a bare "error code: 256" or "error code: 2048".
USDZ files whose material uses both a metalness map and a roughness map now convert to GLB.
USDZ to GLB conversions now keep the original texture colors. Textures had come out too bright after the previous USDZ fix because the colors were corrected twice.
To invite someone to a project, open the share dialog and click Create link. The link appears in a field you can copy.
Invite links are valid for 24 hours, down from 7 days.
Very large DWG and DXF drawings now convert in seconds, not minutes.
Convert3D also frees memory after converting very large models, so big files convert more reliably.
Convert3D now keeps a project's images and converted 3D previews, so reopening a project brings back previews, lightboxes and attachment thumbnails faster.
When Convert3D can't decode 3D solids kept inside blocks in a DWG file, it now says the solid data couldn't be decoded and suggests using Save As in AutoCAD to refresh the file. It used to say the scene was empty or couldn't be tessellated.
The Projects list now has checkboxes, so you can select several projects and delete them together after a confirmation.
A project can hold up to 200 attachments. The upload dialog tells you when you reach the limit and gives an email address to write to if you need more.
Transfers and requests now share one Projects list. You can sort by column, filter by who owns a project, see collaborators and expiry dates, and start a project with the New Project button.
Inside a project, the conversation shows each new version of a model. Resolved comments are marked. Two 3D previews can follow the same camera, so you can compare versions.
Comments, replies, title edits, renamed attachments and resolved threads now appear in a project as they happen, with no reload. The 3D preview stays where you left it.
Sign in now takes you back to the page you opened, such as a shared link or a request.
Request previews also gained attachment previews and a larger image view. The camera and background you pick are remembered.
Large DXF drawings now read in seconds instead of minutes.
SketchUp conversions now keep all their faces, including in large files, and textures land on the right faces.
Some conversions had lost triangles because the reader mistook ordinary data for geometry.
You could import SketchUp files but not export them. Now you can convert 3D models to SKP and open them in SketchUp.
Meshes, materials, textures and UV mapping carry over. Each mesh in your model becomes its own component in SketchUp.
DWG files with 3D solids now convert on the website too, after we updated the part that reads solids. Some had stopped with a message that began "DWG parsed but contained only geometry".
DWG and DXF files with 3D solids from mechanical CAD now convert to solid shapes you can view, including very large assemblies. You can also export a DWG or DXF to STEP.
Convert3D used to ask you to re-export these files as STEP.
3ds Max files now convert with the right vertices, whole meshes and hidden objects staying hidden. V-Ray and Corona materials no longer look see-through, and very large files fit in memory.
Some models can still come out at the wrong size, because the importer doesn't yet read the unit set in the file.
SketchUp import now reads both the older and newer SKP file layouts, and keeps textures.
SketchUp (SKP) files are no longer Pro. You can convert them without a subscription, even when signed out.
DWG and DXF conversions through api.convert3d.org now finish, both when reading and when writing. Some had hung or failed.
Characters with a skeleton keep their size, rigs keep their orientation, and mirror modifiers are applied, so the whole model comes through. Models with millions of vertices also use much less memory.
Some of these files had come out wrong in glTF before.
You could read DXF and DWG files but not write them. Now you can export both.
Models from engineering formats such as SolidWorks and STEP keep exact lines, circles, ellipses and curves. Other 3D models export as meshes. For CAD formats you choose between a 3D export and a 2D drawing with three views.
After a conversion or compression, the sidebar has a Continue in section with Convert, Compress, Render, View and Transfer. Click one and your result moves there, with no download and upload.
If the handoff fails, you see an error and can download and upload the file as before.