What you want to do is highly disruptive if you have published history to other developers. See “Recovering From Upstream Rebase” in the git rebase documentation for the necessary steps after repairing your history.

You have at least two options: git filter-branch and an interactive rebase, both explained below.

Using git filter-branch

I had a similar problem with bulky binary test data from a Subversion import and wrote about removing data from a git repository.

Say your git history is:

$ git lola --name-status
* f772d66 (HEAD, master) Login page
| A     login.html
* cb14efd Remove DVD-rip
| D     oops.iso
* ce36c98 Careless
| A     oops.iso
| A     other.html
* 5af4522 Admin page
| A     admin.html
* e738b63 Index
  A     index.html

Note that git lola is a non-standard but highly useful alias. (See the addendum at the end of this answer for details.) The --name-status switch to git log shows tree modifications associated with each commit.

In the “Careless” commit (whose SHA1 object name is ce36c98) the file oops.iso is the DVD-rip added by accident and removed in the next commit, cb14efd. Using the technique described in the aforementioned blog post, the command to execute is:

git filter-branch --prune-empty -d /dev/shm/scratch \
  --index-filter "git rm --cached -f --ignore-unmatch oops.iso" \
  --tag-name-filter cat -- --all

Options:

• --prune-empty removes commits that become empty (i.e., do not change the tree) as a result of the filter operation. In the typical case, this option produces a cleaner history.
• -d names a temporary directory that does not yet exist to use for building the filtered history. If you are running on a modern Linux distribution, specifying a tree in /dev/shm will result in faster execution.
• --index-filter is the main event and runs against the index at each step in the history. You want to remove oops.iso wherever it is found, but it isn’t present in all commits. The command git rm --cached -f --ignore-unmatch oops.iso deletes the DVD-rip when it is present and does not fail otherwise.
• --tag-name-filter describes how to rewrite tag names. A filter of cat is the identity operation. Your repository, like the sample above, may not have any tags, but I included this option for full generality.
• -- specifies the end of options to git filter-branch
• --all following -- is shorthand for all refs. Your repository, like the sample above, may have only one ref (master), but I included this option for full generality.

After some churning, the history is now:

$ git lola --name-status
* 8e0a11c (HEAD, master) Login page
| A     login.html
* e45ac59 Careless
| A     other.html
|
| * f772d66 (refs/original/refs/heads/master) Login page
| | A   login.html
| * cb14efd Remove DVD-rip
| | D   oops.iso
| * ce36c98 Careless
|/  A   oops.iso
|   A   other.html
|
* 5af4522 Admin page
| A     admin.html
* e738b63 Index
  A     index.html

Notice that the new “Careless” commit adds only other.html and that the “Remove DVD-rip” commit is no longer on the master branch. The branch labeled refs/original/refs/heads/master contains your original commits in case you made a mistake. To remove it, follow the steps in “Checklist for Shrinking a Repository.”

$ git update-ref -d refs/original/refs/heads/master
$ git reflog expire --expire=now --all
$ git gc --prune=now

For a simpler alternative, clone the repository to discard the unwanted bits.

$ cd ~/src
$ mv repo repo.old
$ git clone file:///home/user/src/repo.old repo

Using a file:///... clone URL copies objects rather than creating hardlinks only.

Now your history is:

$ git lola --name-status
* 8e0a11c (HEAD, master) Login page
| A     login.html
* e45ac59 Careless
| A     other.html
* 5af4522 Admin page
| A     admin.html
* e738b63 Index
  A     index.html

The SHA1 object names for the first two commits (“Index” and “Admin page”) stayed the same because the filter operation did not modify those commits. “Careless” lost oops.iso and “Login page” got a new parent, so their SHA1s did change.

Interactive rebase

With a history of:

$ git lola --name-status
* f772d66 (HEAD, master) Login page
| A     login.html
* cb14efd Remove DVD-rip
| D     oops.iso
* ce36c98 Careless
| A     oops.iso
| A     other.html
* 5af4522 Admin page
| A     admin.html
* e738b63 Index
  A     index.html

you want to remove oops.iso from “Careless” as though you never added it, and then “Remove DVD-rip” is useless to you. Thus, our plan going into an interactive rebase is to keep “Admin page,” edit “Careless,” and discard “Remove DVD-rip.”

Running $ git rebase -i 5af4522 starts an editor with the following contents.

pick ce36c98 Careless
pick cb14efd Remove DVD-rip
pick f772d66 Login page

# Rebase 5af4522..f772d66 onto 5af4522
#
# Commands:
#  p, pick = use commit
#  r, reword = use commit, but edit the commit message
#  e, edit = use commit, but stop for amending
#  s, squash = use commit, but meld into previous commit
#  f, fixup = like "squash", but discard this commit's log message
#  x, exec = run command (the rest of the line) using shell
#
# If you remove a line here THAT COMMIT WILL BE LOST.
# However, if you remove everything, the rebase will be aborted.
#

Executing our plan, we modify it to

edit ce36c98 Careless
pick f772d66 Login page

# Rebase 5af4522..f772d66 onto 5af4522
# ...

That is, we delete the line with “Remove DVD-rip” and change the operation on “Careless” to be edit rather than pick.

Save-quitting the editor drops us at a command prompt with the following message.

Stopped at ce36c98... Careless
You can amend the commit now, with

        git commit --amend

Once you are satisfied with your changes, run

        git rebase --continue

As the message tells us, we are on the “Careless” commit we want to edit, so we run two commands.

$ git rm --cached oops.iso
$ git commit --amend -C HEAD
$ git rebase --continue

The first removes the offending file from the index. The second modifies or amends “Careless” to be the updated index and -C HEAD instructs git to reuse the old commit message. Finally, git rebase --continue goes ahead with the rest of the rebase operation.

This gives a history of:

$ git lola --name-status
* 93174be (HEAD, master) Login page
| A     login.html
* a570198 Careless
| A     other.html
* 5af4522 Admin page
| A     admin.html
* e738b63 Index
  A     index.html

which is what you want.

Addendum: Enable git lola via ~/.gitconfig

Quoting Conrad Parker:

The best tip I learned at Scott Chacon’s talk at linux.conf.au 2010, Git Wrangling - Advanced Tips and Tricks was this alias:

lol = log --graph --decorate --pretty=oneline --abbrev-commit

This provides a really nice graph of your tree, showing the branch structure of merges etc. Of course there are really nice GUI tools for showing such graphs, but the advantage of git lol is that it works on a console or over ssh, so it is useful for remote development, or native development on an embedded board …

So, just copy the following into ~/.gitconfig for your full color git lola action:

[alias]
        lol = log --graph --decorate --pretty=oneline --abbrev-commit
        lola = log --graph --decorate --pretty=oneline --abbrev-commit --all
[color]
        branch = auto
        diff = auto
        interactive = auto
        status = auto

(The best answer I've seen to this problem is: https://stackoverflow.com/a/42544963/714112 , copied here since this thread appears high in Google search rankings but that other one doesn't)

A blazingly fast shell one-liner

This shell script displays all blob objects in the repository, sorted from smallest to largest.

For my sample repo, it ran about 100 times faster than the other ones found here.
On my trusty Athlon II X4 system, it handles the Linux Kernel repository with its 5,622,155 objects in just over a minute.

The Base Script

git rev-list --objects --all \
| git cat-file --batch-check='%(objecttype) %(objectname) %(objectsize) %(rest)' \
| awk '/^blob/ {print substr($0,6)}' \
| sort --numeric-sort --key=2 \
| cut --complement --characters=13-40 \
| numfmt --field=2 --to=iec-i --suffix=B --padding=7 --round=nearest

When you run above code, you will get nice human-readable output like this:

...
0d99bb931299  530KiB path/to/some-image.jpg
2ba44098e28f   12MiB path/to/hires-image.png
bd1741ddce0d   63MiB path/to/some-video-1080p.mp4

Fast File Removal

Suppose you then want to remove the files a and b from every commit reachable from HEAD, you can use this command:

git filter-branch --index-filter 'git rm --cached --ignore-unmatch a b' HEAD

100 times faster than git filter-branch and simpler

There are very good answers in this thread, but meanwhile many of them are outdated. Using git-filter-branch is no longer recommended, because it is difficult to use and awfully slow on big repositories.

git-filter-repo is much faster and simpler to use.

git-filter-repo is a Python script, available at github: https://github.com/newren/git-filter-repo . When installed it looks like a regular git command and can be called by git filter-repo.

You need only one file: the Python3 script git-filter-repo. Copy it to a path that is included in the PATH variable. On Windows you may have to change the first line of the script (refer INSTALL.md). You need Python3 installed installed on your system, but this is not a big deal.

First you can run

git filter-repo --analyze

This helps you to determine what to do next.

You can delete your DVD-rip file everywhere:

git filter-repo --invert-paths --path-match DVD-rip
 

Filter-repo is really fast. A task that took around 9 hours on my computer by filter-branch, was completed in 4 minutes by filter-repo. You can do many more nice things with filter-repo. Refer to the documentation for that.

Warning: Do this on a copy of your repository. Many actions of filter-repo cannot be undone. filter-repo will change the commit hashes of all modified commits (of course) and all their descendants down to the last commits!

According to GitHub Documentation, just follow these steps:

1. Get rid of the large file

Option 1: You don't want to keep the large file:

rm path/to/your/large/file        # delete the large file

Option 2: You want to keep the large file into an untracked directory

mkdir large_files                       # create directory large_files
touch .gitignore                        # create .gitignore file if needed
'/large_files/' >> .gitignore           # untrack directory large_files
mv path/to/your/large/file large_files/ # move the large file into the untracked directory

2. Save your changes

git add path/to/your/large/file   # add the deletion to the index
git commit -m 'delete large file' # commit the deletion

3. Remove the large file from all commits

git filter-branch --force --index-filter \
  "git rm --cached --ignore-unmatch path/to/your/large/file" \
  --prune-empty --tag-name-filter cat -- --all
git push <remote> <branch>

This will remove it from your history

git filter-branch --force --index-filter 'git rm -r --cached --ignore-unmatch bigfile.txt' --prune-empty --tag-name-filter cat -- --all

git filter-branch is a powerful command which you can use it to delete a huge file from the commits history. The file will stay for a while and Git will remove it in the next garbage collection. Below is the full process from deleteing files from commit history. For safety, below process runs the commands on a new branch first. If the result is what you needed, then reset it back to the branch you actually want to change.

# Do it in a new testing branch
$ git checkout -b test

# Remove file-name from every commit on the new branch
# --index-filter, rewrite index without checking out
# --cached, remove it from index but not include working tree
# --ignore-unmatch, ignore if files to be removed are absent in a commit
# HEAD, execute the specified command for each commit reached from HEAD by parent link
$ git filter-branch --index-filter 'git rm --cached --ignore-unmatch file-name' HEAD

# The output is OK, reset it to the prior branch master
$ git checkout master
$ git reset --soft test

# Remove test branch
$ git branch -d test

# Push it with force
$ git push --force origin master

Other than git filter-branch (slow but pure git solution) and BFG (easier and very performant), there is also another tool to filter with good performance:

https://github.com/xoofx/git-rocket-filter

From its description:

The purpose of git-rocket-filter is similar to the command git-filter-branch while providing the following unique features:

• Fast rewriting of commits and trees (by an order of x10 to x100).
• Built-in support for both white-listing with --keep (keeps files or directories) and black-listing with --remove options.
• Use of .gitignore like pattern for tree-filtering
• Fast and easy C# Scripting for both commit filtering and tree filtering
• Support for scripting in tree-filtering per file/directory pattern
• Automatically prune empty/unchanged commit, including merge commits

This works perfectly for me : in git extensions :

right click on the selected commit :

reset current branch to here :

hard reset ;

It's surprising nobody else is able to give this simple answer.

git reset --soft HEAD~1

It will keep the changes but remove the commit then you can re-commit those changes.