ClashWise
Local AI clash naming
Turn generic clash names into useful descriptions inside Navisworks. Choose your naming standard and language, then review the proposed titles before applying them. Local naming works offline once set up.
Included in every plan. No per-title charges.
Inside Navisworks Manage 2024, 2025, 2026, 2027With local mode and cloud fallback off, clash content stays on your machine during naming. First sign-in and model download need the internet.
Compare naming examplesLocal AI powers Title Update. Your team can name large batches and refine the wording without consuming Wise actions or paying for each title.
Element names, file names, levels and disciplines stay local during naming when cloud fallback is off. Usage analytics may still include title counts, naming standard, language and timestamp, but not element names, file paths or clash content.
Run a large test, repeat it with another standard or name groups as well as individual clashes. Paid local naming has no per-title quota.
English, German, Spanish, French, Italian, Japanese, Korean, Brazilian Portuguese, Russian, Simplified Chinese and Traditional Chinese. Technical identifiers remain readable. Choose a model that supports your output language.
A Windows computer with a recent CPU and 16 GB RAM is recommended. Allow about 5 GB free disk for the starter model and cache; larger models need more space and memory.
Download the plug-inOpen Title Update or Local AI from the ClashWiseAI ribbon. Download the recommended Qwen 2.5 1.5B model, about 1 GB, then select Use. This first setup needs an internet connection.
Set Name, Element ID, Category, Level and Discipline once per source-file pattern. The mapping is saved per document and can be applied to other models, whether they originate in Revit, Civil 3D, AutoCAD, Tekla, Archicad, IFC or another supported source.
Explore Property MappingChoose tests, naming standard, language and output: title, description or comment. Review is on by default. Edit, reject or regenerate proposals, then apply the accepted titles. Restore the run or use Navisworks Undo if needed.
The curated catalog offers ten models, from about 1 GB to 12 GB. Start small, test on a representative clash, then compare larger options if you need different wording. Larger models use more memory and can run more slowly.
Qwen 2.5 1.5B · Qwen 2.5 3B
Gemma 2 2B · Gemma 3 4B · Gemma 4 E2B · Gemma 4 E4B
Llama 3.2 3B
Phi-3.5 Mini
Mistral 7B v0.3
GPT-OSS 20B
Models are downloaded from public registries and carry their providers’ licence terms. ClashWise does not own the underlying language models.
Pro and Business include both. Starter includes the local toolkit. Cloud title fallback is an optional setting for eligible subscribers; leave it off to keep naming content on your computer.
Manage models and naming settings inside the plug-in, alongside the clash tests you already use.
The AI Models manager shows download progress, integrity checks, supported languages and a recommended model. Generate a test title before using a model for a full batch.
See the active model during generation, control how long it stays in memory and choose whether cloud fallback is allowed. Non-English naming uses a separate translation pass.
Choose individual clashes, groups or both. Keep the original clash or test name as a prefix. Six built-in standards are joined by custom formats; token templates apply exactly as typed, without AI translation.
The local engine runs separately from Navisworks and can be restarted if it fails. It still uses your computer’s CPU and memory, so choose a model suited to your hardware. The current local engine runs on CPU.
The Title Update run sheet keeps test selection beside naming standard, language, Property Mapping and model readiness. Open the dedicated review window to inspect the proposed titles and decide what to apply.
For a batch that also runs tests and publishes results, enable the optional local title stage in Run & Publish. It has no cloud fallback, can pause for review and preserves the current title when generation fails. Local naming and cloud publishing have separate plan requirements.
See Run & PublishQuality depends on the chosen model, source properties and ambiguity of the clash. Larger cloud models can handle difficult wording differently. Try a representative test and review the proposed titles before applying them.
Local inference runs on your machine with cloud fallback off. Title modifications may still sync audit metadata, including the generated title and source filenames. Publishing and cloud collaboration are separate online actions. Read the security page for these data boundaries.
Sign in, download and activate a model, then complete Property Mapping for each source-file pattern. Save the Navisworks document before running. The mapping can be reused on other models.
A recent CPU, 16 GB RAM and about 5 GB free disk are recommended for the starter model and cache. Larger models need more memory and disk. The engine runs on CPU; GPU acceleration is not part of this local workflow.
The recommended Qwen 2.5 1.5B model is about 1 GB. The current catalog reaches about 12 GB for the largest model. Download only the models you need and remove unused ones in AI Models.
The AI runs in a separate process, but shares your computer’s CPU and memory. Speed depends on the model, hardware and batch size. If the engine fails, the plug-in can restart it independently of Navisworks.
Yes, after first sign-in and the one-time model download. Subscription access is re-verified periodically: after up to 14 days offline, you will need to reconnect once before continuing. Dashboard usage sync also needs a connection.
Eleven: English, German, Spanish, French, Italian, Japanese, Korean, Brazilian Portuguese, Russian, Simplified Chinese and Traditional Chinese. The recommended model supports all eleven. Technical identifiers are preserved; custom token templates are not translated.
Local naming is included in Starter, Pro and Business. Starter costs US$19/month or US$190/year and is local-only, with no metered naming usage. Pro and Business add cloud collaboration. You can cancel in the app.
Those use the cloud for shared coordination and require the appropriate plan. Local AI is the title-generation engine in Navisworks; it does not move those collaboration features offline.
No extra charge for Pro or Business subscribers. If you only need the local toolkit, choose Starter. Local AI is also included in the 14-day trial, with no credit card required.
Qwen 2.5 1.5B from Alibaba is the recommended starter model. The AI Models window shows the active model and lets you choose among the ten catalog options, test them, download them or remove them.
Use the models in the supported catalog. Drop-in custom GGUF models are not part of the current workflow.
Try local naming on a small test from your own project. Review the result, then apply the titles you want to keep.