niklaslenzOther tools, one page each
AI tools that fit no text, image or video model page: image to 3D and speech to text so far. Timed on my own machines, every number links to its raw file.
2 tools
| Category | Model | Hardware | Key number | Measured |
|---|---|---|---|---|
| Image to 3D | Idol | RTX 5060 Ti 16 GB and RTX 5090 32 GB | 14.5 s per picture on the RTX 5090, 57.4 s on the RTX 5060 Ti median of 21 pictures at 262,144 Gaussians | 31 Oct 2026 |
| Speech to text | Nemotron 3.5 Transcribe 30B-A4B | RTX 5060 Ti 16 GB | 59:39 of audio in 120 s, 29.8× real time word error rate 1.54 % | 29 Oct 2026 |
The tools are not ranked: they do different jobs, so a number on one row says nothing about the other row. Click a model for the full page with the method, every number linked to its raw file and the exact command.
Image to 3D and speech to text
Idol, image to 3D, on an RTX 5060 Ti (16 GB) and an RTX 5090 (32 GB)
RTX 5060 Ti 16 GB and RTX 5090 32 GB · One picture of an object in, one 3D Gaussian splat (a PLY file) out
Full page: numbers, notes, method · Exact command and run record · All raw files
Nemotron 3.5 Transcribe, speech to text, on an RTX 5060 Ti (16 GB)
RTX 5060 Ti 16 GB · English speech in, text out
Full page: numbers, notes, method · Exact command and run record · All raw files
What is the same for every tool
Each tool gets one fixed test on a free machine, with the exact command, the settings and the raw output published. The time is the wall time of the job, and the page says what is inside it. Nothing here is scored or ranked. Pictures, clips and transcripts on these pages are AI-generated.