We usually just dump a whole load SHOW CREATE TABLE dumps into drawsql at work when we discuss database migrations, and then drop a bunch of "tables" with no columns as comments. I am not aware of an tool that could deal with procedures and such, but we've only looked at that, SQLyog and Intellij's built-in visualizers.

which tables connect to which other tables

You get no guarantees whatsoever that the database schema actually contains this information.

Some databases are set up with validation rules that an external tool can understand, that column FOO.BAR_ID represents a foreign key for column BAR.ID

But some databases (for example mine) have no such rules and depend on the developers to know this through documentation or occult consultation with the spirits of departed coworkers.

A lot of what you're looking for will be in the Data Dictionary Views, been a while since I've worked with oracle, but done something similar for tables and constraints in sqlserver. There's scheduler, plsql, table, constraints (hopefully you have foreign keys) and column information available, a lot more as well.

Oracle SQL developer can import from an existing database into physical and relational models, Erwin and Redgate exist as well. But before going that route, DBAs may very well have the information you're looking for, hopefully it's modeled somewhere.

Dump every single schema, paste the whole thing to claude.ai, ask it to generate graphviz markup showing the whole thing as you requested. Give the result to graphviz, see what happens.

It might not work or might not work perfectly, but that is what I would try and it definitely has the potential to be better than staring at the text trying to make sense of it all. Depending on the size of the beast, you might have to do it in pieces with a defined mapping of table names to graphviz node names so that piecing it out works smoothly.