Reka (https://reka.ai/ and models: https://huggingface.co/RekaAI) offers several multimodal models for image and video analysis, text and voice.
Two first versions were offered here: Reka and Reka’s Space.

In the interface, several possibilities:

• Reka Edge for image and video analysis (categorization and search)
• Reka Research: a web-connected chatbot for in-depth research or interactions on specific documents
• Reka Chat: a chatbot based on the “Reka Flash” model
• Reka Speech: for transcription, translation and generation of voice audios.

Note: other applications are offered outside the main interface, “Clip” (paid) to extract video excerpts and “Vision platform” (90 min of free video) to classify and search in a bank of personal videos.
For Reka Edge, sometimes answers in English.

No pricing page on the site, observations during the tests:

• Works without an account and for free with some limited features
• History is saved with an account
• No limits have been identified on the four main tools.

Some of the models are open-source (see link above).

Images for testing: https://commons.wikimedia.org/wiki/File:Louis-DelphinOdobeyCadet.jpg

Sounds and videos: files proposed for testing.