How Long Does Image to 3D Generation Take?
Real generation times per tier: about 30s on Quick, 60-90s on Quick Plus, 3-5 min on Precise, and why the estimates come from live completion data.
Short answer
Image-to-3D generation on MakeIt3D takes about 30 seconds on the Quick tier, 60 to 90 seconds on Quick Plus, and 3 to 5 minutes on Precise. These are live numbers, not guesses: the app averages the last 20 completed generations per tier (excluding queue time and outliers) and adjusts the estimate for current load, adding roughly 8 percent per concurrent generation up to a 1.5x cap, plus a 30-second cold-start penalty if that tier has not run in the past hour. The tiers differ because they run different AI models. Quick uses TRELLIS (1 credit), tuned for speed with a 1,024-pixel texture. Quick Plus runs TRELLIS 2 (1 credit) with a 2,048-pixel texture and remeshing, roughly doubling the wait. Precise runs Hyper3D Rodin v2 (3 credits), which builds PBR materials and up to 500,000 polygons; that deeper reconstruction is why it takes minutes instead of seconds.
What happens during those seconds
A generation is more than one AI call. Here is the actual pipeline and where the time goes at each tier.
Upload and preprocessing
Your photo is uploaded and passed through background removal before reconstruction, so the AI works from a clean object silhouette rather than the whole scene. This preprocessing adds a few seconds and noticeably improves geometry.
The estimate you see is measured, not fixed
The wait-time endpoint averages the last 20 completed generations for your chosen tier, subtracts recorded queue time, and discards outliers (anything over 2 minutes for Quick, 2.5 for Quick Plus, or 15 for Precise). It then applies a load multiplier of about 8 percent per active generation, capped at 1.5x, plus 30 seconds if the tier is cold.
Quick: about 30 seconds
Quick runs the TRELLIS model for 1 credit with a 1,024-pixel texture and light built-in mesh simplification. It is the iteration tier: fast enough to test several photos or crops before committing to a final render.
Quick Plus: about 60 to 90 seconds
Quick Plus runs TRELLIS 2 for the same 1 credit but does more work: a 2,048-pixel texture, remeshing for cleaner topology, and a decimation target of 500,000 triangles. Expect roughly double the Quick wait.
Precise: 3 to 5 minutes
Precise runs Hyper3D Rodin v2 for 3 credits. It reconstructs up to 500,000 polygons with PBR materials, which involves substantially deeper geometry and texture passes. Use it for final assets and prints, not for exploration.
Delivery and download
The finished model streams into the browser viewer as a GLB. Converting to STL, OBJ, or 3MF happens in your browser at download time, so switching formats adds seconds, not another generation wait.
FAQ
Why does Precise take up to ten times longer than Quick?
It runs a heavier model. Hyper3D Rodin v2 builds up to 500,000 polygons with PBR materials, while Quick's TRELLIS pass optimizes for speed with a smaller texture. The output difference is visible on faces, fine ornament, and print detail.
Do multiple photos make generation slower?
Not meaningfully. A generation with up to 4 photos runs in roughly the same time band as a single-photo generation on the same tier, and it costs the same number of credits.
Why did the estimated wait change between visits?
The estimate is recomputed from the most recent completed generations and current load. Busy periods push it up by as much as 50 percent, and the first generation after a quiet hour can include a roughly 30-second model warm-up.
What is the fastest way to iterate on a model?
Test photos and crops on Quick at 1 credit and about 30 seconds each, then rerun only the winning photo on Precise. This costs less and is much faster than iterating directly on the 3-to-5-minute tier.
Does downloading in a different format add wait time?
Only seconds. The model is stored as GLB and converted to STL, OBJ, or 3MF locally in your browser when you download, with no server round trip or credit cost.