SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World
Back to feed
Programming

ComfyUI TRELLIS.2 Test Cuts a 1.18M-Triangle Mesh to a 19,738-Triangle Prop

0
·3 views

A local ComfyUI workflow using TRELLIS.2 generated a raw mesh of 1,175,882 triangles from an official example image. Remeshing, decimation, UV unwrapping and texture baking reduced the deliverable to 19,738 triangles in a 4.55 MB textured GLB. Blender and Houdini imports showed that an embedded texture working in Blender was not directly usable by the Houdini SOP's generated material. The test ran on an RTX 5070 Ti with 16 GB VRAM, using Blender 5.2.0 LTS and Houdini 22.0.429 for the DCC checks. The author omitted the high-resolution upsample branch for this 16 GB setup, so the result is not the unmodified highest-resolution template output.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Palo Alto PAN-OS flaw CVE-2026-0321 allows unauthenticated denial-of-service via dataplane traffic

Palo Alto Networks released a batch of security advisories in mid-October 2026, which CERT-FR grouped under advisory CERTFR-2026-AVI-1326. Most issues in the batch require local or administrative access, but CVE-2026-0321 stands apart because an unauthenticated attacker with network reach can send crafted traffic to a dataplane interface. Repeated exploitation can trigger a denial-of-service state that pushes the device into maintenance mode. The flaw is rated MEDIUM at 6.6 under CVSS 4.0, reflecting its availability-only impact rather than the ease of exploitation. Analysts recommend handling it on a separate, shorter timeline from the rest of the batch, since its effect is on traffic rather than credentials.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why PostgreSQL VACUUM Monitoring Matters for Preventing Table Bloat

Table bloat in PostgreSQL can degrade performance for weeks before it is noticed, since symptoms like growing disk usage and slower queries appear indirectly. The cause lies in MVCC, where updates and deletes leave dead row versions that only VACUUM can reclaim, and only once no running transaction still needs them. Autovacuum may be enabled yet still fall behind, because its default trigger thresholds scale poorly on very large tables. Long-running transactions, idle sessions, and replication slots can also hold back cleanup, while cost-limited workers may never catch up on write-heavy tables. Effective monitoring of dead tuples and vacuum activity is therefore needed to catch these problems early.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Only 53 of 1,045 Hours of Sleep Audio Were Usable for Training a Sleep-Phase Model

A developer team recorded about 1,045 hours of overnight bedroom audio to train a sleep-phase model, but only roughly 53 hours met the criteria for training. The main obstacle was not capturing audio, which they described as a solved problem, but obtaining reliable labels that mark sleep stages. Their labels came from wearable devices, which only worked on nights when the phone and wearable recordings overlapped cleanly, and small clock drift or early device failures removed usable hours. They concluded that defining what counts as a valid night, and discarding incomplete recordings, was more important than collecting more data. Their practices included logging each recording's completeness, aligning audio to labels before labelling, and tracking where each label came from.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Tool-cost CI check for MCP servers catches schema growth hidden by token totals

A developer guide describes a reproducible CI check for MCP tool definitions that measures total tokens, descriptions and input schemas separately. It uses an observed remote change to textokit's endpoint, where a 20-tool list lost about 32% of its counted tokens overall, mostly from shorter descriptions. The article provides a full fetch script that verifies canonical digests of the public tool lists before comparing them. Commands use pinned versions of the aimarket-histor CLI and tiktoken, and the author notes this is their own reproducibility check rather than external adoption. The check exits with code 0 when the increase stays within a configured threshold of 10%.

0 comments Read more at DEV Community