diff --git a/seeker.go b/seeker.go
new file mode 100644
index 0000000..3c164e3
--- /dev/null
+++ b/seeker.go
@@ -0,0 +1,80 @@
+package sfstreams
+
+import (
+	"io"
+	"sync"
+)
+
+type parentSeeker struct {
+	io.ReadSeekCloser
+	underlying io.ReadSeekCloser
+	mutex      *sync.Mutex
+	closeWg    *sync.WaitGroup
+}
+
+func newParentSeeker(src io.ReadSeekCloser, downstreamReaders int) *parentSeeker {
+	wg := new(sync.WaitGroup)
+	wg.Add(downstreamReaders)
+	go func() {
+		wg.Wait()
+		_ = src.Close()
+	}()
+	return &parentSeeker{
+		underlying: src,
+		mutex:      new(sync.Mutex),
+		closeWg:    wg,
+	}
+}
+
+func (p *parentSeeker) Read(b []byte) (int, error) {
+	return p.underlying.Read(b)
+}
+
+func (p *parentSeeker) Seek(offset int64, whence int) (int64, error) {
+	return p.underlying.Seek(offset, whence)
+}
+
+func (p *parentSeeker) Close() error {
+	return p.underlying.Close()
+}
+
+type downstreamSeeker struct {
+	io.ReadSeekCloser
+	parent *parentSeeker
+	pos    int64
+}
+
+func newSyncSeeker(parent *parentSeeker) *downstreamSeeker {
+	return &downstreamSeeker{
+		parent: parent,
+		pos:    0,
+	}
+}
+
+func (s *downstreamSeeker) Read(b []byte) (int, error) {
+	s.parent.mutex.Lock()
+	defer s.parent.mutex.Unlock()
+	offset, err := s.parent.Seek(s.pos, io.SeekStart)
+	if err != nil {
+		return 0, err
+	}
+	i, err := s.parent.Read(b)
+	s.pos = offset + int64(i)
+	return i, err
+}
+
+func (s *downstreamSeeker) Seek(offset int64, whence int) (int64, error) {
+	s.parent.mutex.Lock()
+	defer s.parent.mutex.Unlock()
+	offset, err := s.parent.Seek(offset, whence)
+	if err != nil {
+		return s.pos, err
+	}
+	s.pos = offset
+	return s.pos, nil
+}
+
+func (s *downstreamSeeker) Close() error {
+	s.parent.closeWg.Done()
+	return nil
+}
diff --git a/sfstreams.go b/sfstreams.go
index df4d083..4a560b9 100644
--- a/sfstreams.go
+++ b/sfstreams.go
@@ -22,6 +22,17 @@ type Group struct {
 	sf    singleflight.Group
 	mu    sync.Mutex
 	calls map[string][]chan<- io.ReadCloser
+
+	// Normally, the Group will copy the work function's returned reader, but in some cases it is
+	// desirable to maintain the io.Seeker interface. When this option is set to true, the Group
+	// no longer copies but instead returns proxy io.ReadSeekCloser readers that track their own
+	// read state. Whenever one of those readers seeks/reads, it synchronously does so on the work
+	// function's returned io.ReadSeekCloser. This can lead to performance bottlenecks if several
+	// call sites are attempting to read/seek at the same time.
+	//
+	// If this is set to true, but the work function doesn't return an io.ReadSeekCloser, the copy
+	// behaviour is used. When false (the default), the copy behaviour is always used.
+	UseSeekers bool
 }
 
 // Do behaves just like singleflight.Group, with the added guarantee that the returned io.ReadCloser
@@ -92,32 +103,46 @@ func (g *Group) doWork(key string, fn func() (io.ReadCloser, error)) func() (int
 		} else {
 			var zero io.ReadCloser
 			canStream := fnRes != nil && fnRes != zero
-			writers := make([]*io.PipeWriter, 0)
-			for _, ch := range chans {
-				if !canStream {
+
+			delete(g.calls, key) // we've done all we can for this call: clear it before we unlock
+
+			if !canStream {
+				for _, ch := range chans {
 					// This needs to be async to prevent a deadlock
 					go func(ch chan<- io.ReadCloser) {
 						ch <- nil
 					}(ch)
-					continue
 				}
+				return nil, fnErr // we intentionally discard the return value
+			}
 
-				r, w := io.Pipe()
-				writers = append(writers, w)
+			if g.UseSeekers {
+				if rsc, ok := fnRes.(io.ReadSeekCloser); ok {
+					parent := newParentSeeker(rsc, len(chans))
+					for _, ch := range chans {
+						// This needs to be async to prevent a deadlock
+						go func(ch chan<- io.ReadCloser) {
+							ch <- newSyncSeeker(parent)
+						}(ch)
+					}
+				}
+			} else {
+				writers := make([]*io.PipeWriter, len(chans))
+				for i, ch := range chans {
+					r, w := io.Pipe()
+					writers[i] = w
 
-				// This needs to be async to prevent a deadlock
-				go func(r io.ReadCloser, ch chan<- io.ReadCloser) {
-					ch <- newDiscardCloser(r)
-				}(r, ch)
-			}
-			delete(g.calls, key) // we've done all we can for this call: clear it before we unlock
+					// This needs to be async to prevent a deadlock
+					go func(r io.ReadCloser, ch chan<- io.ReadCloser) {
+						ch <- newDiscardCloser(r)
+					}(r, ch)
+				}
 
-			if canStream {
 				// Do the io copy async to prevent holding up other singleflight calls
 				go finishCopy(writers, fnRes)
 			}
 
-			return nil, fnErr // we discard the return value
+			return nil, fnErr // we intentionally discard the return value
 		}
 	}
 }
diff --git a/sfstreams_test.go b/sfstreams_test.go
index a2a63a1..30e7fd8 100644
--- a/sfstreams_test.go
+++ b/sfstreams_test.go
@@ -10,12 +10,24 @@ import (
 	"time"
 )
 
+type nopReadSeekCloser struct {
+	io.ReadSeeker
+}
+
+func (nopReadSeekCloser) Close() error {
+	return nil
+}
+
+func nopSeekCloser(rs io.ReadSeeker) io.ReadSeekCloser {
+	return nopReadSeekCloser{ReadSeeker: rs}
+}
+
 func makeStream() (key string, expectedBytes int64, src io.ReadCloser) {
 	key = "fake file"
 
 	b := make([]byte, 16*1024) // 16kb
 	_, _ = rand.Read(b)
-	src = io.NopCloser(bytes.NewBuffer(b))
+	src = nopSeekCloser(bytes.NewReader(b))
 
 	expectedBytes = int64(len(b))
 	return
@@ -93,7 +105,7 @@ func TestDoDuplicates(t *testing.T) {
 		return src, nil
 	}
 
-	g := &Group{}
+	g := new(Group)
 	readFn := func() {
 		defer workWg2.Done()
 		workWg1.Done()
@@ -323,7 +335,7 @@ func TestStallOnRead(t *testing.T) {
 		return src, nil
 	}
 
-	g := &Group{}
+	g := new(Group)
 	readFn := func(i int) {
 		defer workWg2.Done()
 		workWg1.Done()
@@ -386,7 +398,7 @@ func TestFasterRead(t *testing.T) {
 		return src, nil
 	}
 
-	g := &Group{}
+	g := new(Group)
 	readFn := func(i int) {
 		defer workWg2.Done()
 		workWg1.Done()
@@ -437,3 +449,142 @@ func TestFasterRead(t *testing.T) {
 		t.Errorf("Expected between 1 and %d calls, got %d", max-1, callCount)
 	}
 }
+
+func TestReturnsNoSeekerDefault(t *testing.T) {
+	key, expectedBytes, src := makeStream()
+
+	callCount := 0
+	workFn := func() (io.ReadCloser, error) {
+		callCount++
+		return src, nil
+	}
+
+	g := new(Group)
+	r, err, shared := g.Do(key, workFn)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if shared {
+		t.Error("Expected a non-shared result")
+	}
+	if r == src {
+		t.Error("Reader and source are the same")
+	}
+	if _, ok := r.(io.ReadSeekCloser); ok {
+		t.Error("Expected reader to *not* be a ReadSeekCloser")
+	}
+
+	//goland:noinspection GoUnhandledErrorResult
+	defer r.Close()
+	c, _ := io.Copy(io.Discard, r)
+	if c != expectedBytes {
+		t.Errorf("Read %d bytes but expected %d", c, expectedBytes)
+	}
+
+	if callCount != 1 {
+		t.Errorf("Expected 1 call, got %d", callCount)
+	}
+}
+
+func TestReturnsSeekerWhenEnabled(t *testing.T) {
+	key, expectedBytes, src := makeStream()
+
+	callCount := 0
+	workFn := func() (io.ReadCloser, error) {
+		callCount++
+		return src, nil
+	}
+
+	g := new(Group)
+	g.UseSeekers = true
+	r, err, shared := g.Do(key, workFn)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if shared {
+		t.Error("Expected a non-shared result")
+	}
+	if r == src {
+		t.Error("Reader and source are the same")
+	}
+	if _, ok := r.(io.ReadSeekCloser); !ok {
+		t.Error("Expected reader to be a ReadSeekCloser")
+	}
+
+	//goland:noinspection GoUnhandledErrorResult
+	defer r.Close()
+	c, _ := io.Copy(io.Discard, r)
+	if c != expectedBytes {
+		t.Errorf("Read %d bytes but expected %d", c, expectedBytes)
+	}
+
+	if callCount != 1 {
+		t.Errorf("Expected 1 call, got %d", callCount)
+	}
+}
+
+func TestSeekerUsesParent(t *testing.T) {
+	key, expectedBytes, src := makeStream()
+
+	skip := int64(10)
+
+	workWg1 := new(sync.WaitGroup)
+	workWg2 := new(sync.WaitGroup)
+	workCh := make(chan int, 1)
+	callCount := 0
+	workFn := func() (io.ReadCloser, error) {
+		callCount++
+		if callCount == 1 {
+			workWg1.Done()
+		}
+		v := <-workCh
+		workCh <- v
+		time.Sleep(10 * time.Millisecond)
+		return src, nil
+	}
+
+	g := new(Group)
+	g.UseSeekers = true
+	readFn := func(i int) {
+		localSkip := skip + int64(i)
+		defer workWg2.Done()
+		workWg1.Done()
+		r, err, _ := g.Do(key, workFn)
+		if err != nil {
+			t.Error(err)
+			return
+		}
+		rsc := r.(io.ReadSeekCloser)
+		offset, err := rsc.Seek(localSkip, io.SeekStart)
+		if err != nil {
+			t.Error(err)
+			return
+		}
+		if offset != localSkip {
+			t.Errorf("Expected seek to %d instead of %d", localSkip, offset)
+		}
+		c, err := io.Copy(io.Discard, rsc)
+		if err != nil {
+			t.Error(err)
+			return
+		}
+		if c != (expectedBytes - localSkip) {
+			t.Errorf("Read %d bytes instead of %d", c, expectedBytes)
+		}
+	}
+
+	const max = 10
+	workWg1.Add(1)
+	for i := 0; i < max; i++ {
+		workWg1.Add(1)
+		workWg2.Add(1)
+		go readFn(i)
+	}
+	workWg1.Wait()
+	workCh <- 1
+	workWg2.Wait()
+	if callCount <= 0 || callCount >= max {
+		t.Errorf("Expected between 1 and %d calls, got %d", max-1, callCount)
+	}
+
+}