Bagel (demo: https://demo.bagel-ai.org/ and documentation: https://github.com/ByteDance-Seed/Bagel) is a multimodal model open-source ByteDance including text, image and reasoning.

French is accepted for instructions, but the responses in the text module are not optimized for French.
The image module accepts iteration and simple text (see tests).
Editable settings.

Free, without an account and unlimited.

Image for test: https://fr.wikipedia.org/wiki/Citadelle_de_Besan%C3%A7on#/media/Fichier:Citadelle_Besan%C3%A7on.jpg