Multi-Modal Chat
Text, image, file reading, file generation and voice -- all in one chat window. No tool switching, no data risk.
Five modalities in one chat
Text
Standard chat with all 100+ models, including streaming responses.
Image understanding
Upload screenshots, photos, diagrams -- the AI analyses and answers questions about them.
File reading
PDF, Word, Excel, CSV, TXT -- directly in the chat context, no pre-processing needed.
File generation
Say 'Create PDF with ...', 'Make me Excel with ...' -- HOVIGuard generates the ready file with tables and charts directly in the chat.
Voice input
Speak instead of type -- dictation in 30+ languages with automatic transcription.
Say what you need. Get the file.
Instead of copying an answer and rebuilding it in Word, the multi-modal chat returns a ready file. PDF, Word or Excel -- with tables, research data and charts, embedded.
- Three formats: PDF, DOCX (Word), XLSX (Excel) -- automatically matched to the request
- Charts in the document -- from data the model researches itself
- File is stored encrypted in EU storage and automatically deleted after 24 hours
- No extra tool, no tab switching -- all in the normal chat window
How the multi-modal chat works
Five steps from model choice to answer -- every request secured by the MultiLayer Data Shield.
- 1
Pick a model or hand to Pilot
Choose from 100+ EU models or let Model Pilot route automatically. Code, vision, reasoning, web search -- every mode has the right model.
- 2
Input -- text, image, file or voice
Type, paste, drop a file or use the microphone. All modalities land in the same chat context, no tool switching.
- 3
MultiLayer Data Shield filters before sending
PII detection, brand check, manipulation protection, content filter -- four layers check the input including embedded images and documents.
- 4
Response streams back -- with audit entry
Token by token, with sources where available. Every request automatically receives an audit entry incl. model, cost, protective measures.
- 5
Switch model mid-conversation
Within the same chat you can switch model per request -- Claude for reasoning, Gemini for vision, GPT for creative writing.
Who benefits from the multi-modal chat?
Four roles HOVIGuard relieves daily -- with concrete tasks per use case.
Sales & customer success
Analyse customer requests with attachments, extract briefings from PDFs, multilingual replies in seconds.
Marketing
Image descriptions for SEO and alt tags, tone checks on texts, competitor analyses from screenshots.
IT & support
Log file and error screenshot analysis, documentation search, technical instructions summarised from manuals.
Legal & compliance
Contract review with markup, pre-formulating GDPR requests, risk assessment with citable audit entries.
Benefits for your organisation
Protected by the MultiLayer Data Shield
Every input is checked against four protection layers before being sent to the model -- including images and documents.
View layers →Frequently asked questions on the multi-modal chat
Answers on models, file formats, data protection and technical limits.
Which AI models are available in the multi-modal chat?+
Over 100 EU-hosted models including Claude (Sonnet, Opus, Haiku), GPT family, Google Gemini, Mistral, Meta Llama, Cohere, DeepSeek and xAI Grok. Selectable per request or routed automatically by Model Pilot.
Which file formats can I upload?+
PDF, Word (.docx), Excel (.xlsx), CSV, TXT, Markdown plus images as JPG, PNG, WEBP, HEIC. Max file size 25 MB per upload, up to 5 attachments per request.
Can the chat generate a file for me?+
Yes. Requests like „Create a PDF with ...“, „Make me Excel with ...“ or „Generate a Word document about ...“ are auto-detected. HOVIGuard researches (web search if needed), structures the content, embeds tables and charts, and provides a ready PDF/DOCX/XLSX file for download. The file is stored encrypted in EU storage and automatically deleted after 24 hours.
Is my data used to train the models?+
No. All integrated providers have contractually committed that neither prompts nor outputs are used for model training. The agreements are documented in the data processing addendum (DPA).
How much context (tokens) can the chat process?+
Depending on the model between 32k tokens (smaller models) and 1 million tokens (Claude Opus 4.7 with 1M context, Gemini 2.5). The selection menu shows the limit per model visibly.
How secure is image recognition with personal data?+
Before sending, each image passes through the MultiLayer Data Shield -- faces, ID cards, licence plates are detected and depending on policy masked, blocked or queried back. Configuration per tenant by the company admin.
Can I switch models within the same chat?+
Yes. Pick a different model per request -- e.g. Claude for reasoning, Gemini for vision, GPT for creative writing. The chat history is preserved, the audit log documents every switch.
Where is my data stored?+
Data does not leave the EU. No data transfer to third countries, no sub-processing outside the EU.
How does voice input work?+
Activate the microphone in the chat window, speak, done. Transcription runs locally in the browser or via EU-hosted Whisper instances. Supports 30+ languages including German, English, French, Italian.