This may take several iterations to get complete and correct.
There are two aspects: What happens when you are face to face with students. What happens when students study/learn on their own.
What happens when you are face to face with students.
First, I suggest that when you have students together with you (and staff) that you do only those things that can't be done otherwise. Even some discussion can be done otherwise if you use a chatroom or similar tool.
The typical flipped classroom uses the Face Time (FT) for student practice under observation. One way is to hand out worksheets at the start and have students work on them as you (and staff) wander around looking for blocked students. In CS you can have programming exercises, in mathematics you can have a variety of problems of differing difficulty. In a writing class you can give a writing exercise, in a history class some online research problems (caveat: not my field).
One effective way to manage such a "learning lab" is to have students work in pairs (and rotate the pairs). This results in fewer blockages. If students work alone then when they get blocked they stay blocked until someone on staff gets released from another conversation to help them. Pairs can help one another since two students seldom get blocked at the same point. You can even permit a pair to turn to an adjacent pair for help if no staff is available at that moment.
But I wouldn't "waste" such time on lecture - broadcasting. An exception might be made for a short "hints" lecture if a lot of students are seemingly blocked. Nearly everything you want to say (and write on a whiteboard) during a lecture is available from other sources except in very advanced or specialized courses. Students need practice and feedback. The FT is used to provide both of these.
It is possible, though sometimes difficult to arrange, to encourage the stronger students to serve as tutors for their weaker classmates in a pair. This works well enough as small scale when you know the students, including something of their skills and personalities, but becomes riskier at large scale. Watching someone work isn't the same as working, of course and it is useful for students to struggle with hard problems, perhaps especially in math. This is one reason to rotate pairs. And you have to observe interactions and possibly intervene. But at least you have the opportunity for this when they work under your view rather than when they work alone or in groups you aren't aware of.
What happens when students study/learn on their own.
The other aspect of learning is the delivery of the material for the course. In a flipped classroom much/most/all of this is done when the instructor isn't co-located with the students. There are a variety of ways to provide this and I strongly suggest that you use many of them, not just video. Textbooks are still useful, even if you also use lecture or video or whatever.
But the purpose of this "between FT sessions" work is to prepare them to be effective later during FT.
You can expand and refine your lecture notes and give them to students directly. You will get feedback from them on the quality when they ask questions. You will eventually have your own textbook for the course if you do this consistently and intelligently, avoiding plagiarizing the works of others. You can provide simple exercises that will be expanded on and deepened later during FT.
You can provide lists of internet resources for many courses. In math and CS, for example, wikipedia is very reliable nearly always. The errors that appear there tend to disappear quickly, especially for non-controversial subjects.
If you make videos, I suggest that you make a lot of very short ones. They should be about 10 minutes (ideal) but not more than 15. The should be subtitled if at all possible and should come with a transcript. The latter is easy as you need the transcript (notes) to create it in the first place. A longer topic is covered by a series of short videos. Make an index available.
Short videos have two advantages. For the student, they are less likely to become disinterested or lose their train of thought. You can also provide a few solidifying exercises with each such video segment to firm up the main idea. Students should still be encouraged (required) to take their own notes as this is an important element of active learning. The advantage for the instructor is that a poor segment or one with errors is easier to replace than part of a longer one.
The transcript of a video can be annotated with time marks for important points so that students can "rewind" to something important easily. It can also be annotated with "links" to text or online resources that expand on the current idea.
Traditionally a course is a certain mixture of time spent with an instructor (presentation) and time spent elsewhere but working effectively. In the US the tradition has been a 1:2 ratio. If a course meets for an hour then the student is expected to work an additional two hours. The ratio and the total don't need to change, though you have a bit more flexibility in a flipped classroom.
Say you have an hour for FT. You can then require two hours outside. About an hour of that can be things like video or book/notes study and about another hour can be spent on traditional homework. The ratio of presentation to study needn't change, not the ratio of in class to out of class time.
But it is probably a mistake to try to compress too much presentation into too short a time, even with 10 minute videos. The course is spread out over many weeks traditionally for an educational reason. Presentation needs to be interspersed with practice and feedback. It is just that you are personally present for part of the practice and less (or none) of the presentation.
Making it work.
You ask for ideas about guaranteeing that the students do the non FT work required. One way (they will hate this) is to require them to keep a notebook and to bring it to class. The notebook is subject to inspection by staff. Since you observe the students, and assuming the scale isn't impossible, you get a sense of who is learning and who isn't. For the latter group, a look at their notebook might tell you a lot, not only about their dedication, but also about their misconceptions as they occur. The notebook is the "ticket" to the FT session.
In a traditional classroom lecture, looking over your shoulder makes it easy to see who is taking notes and who isn't. But in that scenario you also don't know if they are putting required time in outside class, so you require things to be handed in. In a flipped classroom that is also flipped. It is easy to see them getting the exercise but hard to observe them following the material. So you need procedures to check on what you can't observe - hence a notebook open to inspection or some similar thing.
Another check on students doing the out of class work is a very short "pop" quiz at the beginning of each FT session.
And it is also useful to have students hand in index cards at the start and/or the end of such a session that contain questions for you.
Beyond the flipped classroom
It is possible to make learning continuous without breaks. Every day, any time of day. One way to do this is with a simple two way asynchronous communication tool. Students can ask questions whenever they arise. You, the staff, and every student see the questions. Anyone, including students can answer the questions. You will find this reduces staff time somewhat, though you have to read the whole stream. You only need to reply when no one else does or someone replies incorrectly.
You need some rules for this of course. No personal questions (grades,...) are asked on such a list and in some classes you don't want to permit solutions to be posted other than by staff.