# CamReport optional models

ONNX conversions used by the CamReport desktop app, downloaded on request (Settings → Smart search /
Ask in your own words / Voice questions). They are unchanged in behaviour from the original models; only the file format and
weight precision differ.

| Folder | Original model | Licence | What changed |
|---|---|---|---|
| `smart_search/` | [google/siglip2-base-patch16-224](https://huggingface.co/google/siglip2-base-patch16-224) | Apache-2.0 | exported to ONNX; text model word table in INT8 |
| `ask/` | [Qwen/Qwen3-1.7B](https://huggingface.co/Qwen/Qwen3-1.7B) | Apache-2.0 | INT4 ONNX Runtime GenAI build for CPUs |
| `voice/` | [openai/whisper-small](https://huggingface.co/openai/whisper-small) | MIT (OpenAI's repository: github.com/openai/whisper); Apache-2.0 on its Hugging Face card | INT8 ONNX export by [onnx-community/whisper-small](https://huggingface.co/onnx-community/whisper-small), unchanged |

Each folder is distributed under the licence of its original model. Apache-2.0: https://www.apache.org/licenses/LICENSE-2.0.
MIT: https://github.com/openai/whisper/blob/main/LICENSE, Copyright (c) 2022 OpenAI. Copyright of the original weights stays
with Google LLC, the Qwen team (Alibaba Cloud) and OpenAI.

## Files and SHA-256

| File | Size | SHA-256 |
|---|---|---|
| `smart_search/siglip2_base_vision.onnx` | 371.8 MB | `1d843587dabbaa1b101b1a0cff926f8fccf2077a57374d74677cfc23414958ca` |
| `smart_search/siglip2_base_text_qemb.onnx` | 539.5 MB | `cce381eada35213b80296f91b960879f9a3856300abd90fb999354ae11fc72b2` |
| `smart_search/siglip2_tokenizer.json` | 34.4 MB | `cb9140fae3ac5122c972d37adf83e1248471a38147ad76f8215c8872c6fd8322` |
| `ask/chat_template.jinja` | 0.0 MB | `51fa65c79bb57f058dc7ef8734884bd325fe9d45bb03a61bfef59785d3bc2da9` |
| `ask/genai_config.json` | 0.0 MB | `93be8ab220c4fe1c073208ec85ee6321a26f2083e439f3cf9cb4eb9ce3cffdb6` |
| `ask/model.onnx` | 0.3 MB | `ed22f07a42ed7c607684f51708b566eae9ee3e08c17f672eaa05de6b02735622` |
| `ask/model.onnx.data` | 1096.7 MB | `f05debfd4d42c7f416f088d3cfc1c8ef1e016898eafee8db15e55a5f0fa32e7c` |
| `ask/tokenizer.json` | 11.4 MB | `be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506` |
| `ask/tokenizer_config.json` | 0.0 MB | `70dbff0c676db3313a55d21970d115b24a9bbc3f8bf7682c7d73cc948c7f7344` |
| `voice/encoder_model_int8.onnx` | 92.3 MB | `2601c9eb2d345c5916d4576d36f663a7c96589740fb2273828c48c3fc2c7db75` |
| `voice/decoder_model_merged_int8.onnx` | 156.8 MB | `ec07c3cbb64172c39791e26ee870a65ac22b458c36722bfe2776b3dbf741e0c9` |
| `voice/tokenizer.json` | 2.5 MB | `27fc476bfe7f17299480be2273fc0608e4d5a99aba2ab5dec5374b4482d1a566` |
| `voice/generation_config.json` | 0.0 MB | `f538b28220c6a6d6f1af1458d4141cacb4ef4963df3de98a19490440c412ddf0` |
| `voice/config.json` | 0.0 MB | `457854d452f17661e197d74aee12b8e74fb75ba30ebfaa7426d0d61ea1e08a18` |