diff --git a/pkg/obiformats/fastseq_json_header.go b/pkg/obiformats/fastseq_json_header.go
index af8de1c..3928613 100644
--- a/pkg/obiformats/fastseq_json_header.go
+++ b/pkg/obiformats/fastseq_json_header.go
@@ -199,8 +199,111 @@ func _parse_json_array_interface(str []byte) ([]interface{}, error) {
 	return values, nil
 }
 
-func _parse_json_header_(header string, sequence *obiseq.BioSequence) string {
+// _parse_json_annotation_field parses a single key/value pair coming from a
+// JSON object (either a FASTA/FASTQ inline JSON header, or the "annotations"
+// field of a JSON sequence record) and applies it to the sequence, special
+// casing the well-known OBITools attributes (id, definition, count, taxid,
+// obiclean_*, merged_*).
+func _parse_json_annotation_field(key []byte, value []byte, dataType jsonparser.ValueType, sequence *obiseq.BioSequence) error {
 	annotations := sequence.Annotations()
+	var err error
+
+	skey := obiutils.UnsafeString(key)
+
+	switch {
+	case skey == "id":
+		sequence.SetId(string(value))
+	case skey == "definition":
+		sequence.SetDefinition(string(value))
+
+	case skey == "count":
+		if dataType != jsonparser.Number {
+			log.Fatalf("%s: Count attribut must be numeric: %s", sequence.Id(), string(value))
+		}
+		count, err := jsonparser.ParseInt(value)
+		if err != nil {
+			log.Fatalf("%s: Cannot parse count %s", sequence.Id(), string(value))
+		}
+		sequence.SetCount(int(count))
+
+	case skey == "obiclean_weight":
+		weight, err := _parse_json_map_int(value)
+		if err != nil {
+			log.Fatalf("%s: Cannot parse obiclean weight %s", sequence.Id(), string(value))
+		}
+		annotations[skey] = weight
+
+	case skey == "obiclean_status":
+		status, err := _parse_json_map_string(value)
+		if err != nil {
+			log.Fatalf("%s: Cannot parse obiclean status %s", sequence.Id(), string(value))
+		}
+		annotations[skey] = status
+
+	case strings.HasPrefix(skey, "merged_"):
+		if dataType == jsonparser.Object {
+			data, err := _parse_json_map_int(value)
+			if err != nil {
+				log.Fatalf("%s: Cannot parse merged slot %s: %v", sequence.Id(), skey, err)
+			} else {
+				annotations[skey] = obiseq.MapAsStatsOnValues(data)
+			}
+		} else {
+			log.Fatalf("%s: Cannot parse merged slot %s", sequence.Id(), skey)
+		}
+
+	case skey == "taxid":
+		if dataType == jsonparser.Number || dataType == jsonparser.String {
+			taxid := string(value)
+			sequence.SetTaxid(taxid)
+		} else {
+			log.Fatalf("%s: Cannot parse taxid %s", sequence.Id(), string(value))
+		}
+
+	case strings.HasSuffix(skey, "_taxid"):
+		if dataType == jsonparser.Number || dataType == jsonparser.String {
+			rank := skey[:len(skey)-len("_taxid")]
+
+			taxid := string(value)
+			sequence.SetTaxid(taxid, rank)
+		} else {
+			log.Fatalf("%s: Cannot parse taxid %s", sequence.Id(), string(value))
+		}
+
+	default:
+		skey = strings.Clone(skey)
+		switch dataType {
+		case jsonparser.String:
+			annotations[skey] = string(value)
+		case jsonparser.Number:
+			// Try to parse the number as an int at first then as float if that fails.
+			annotations[skey], err = jsonparser.ParseInt(value)
+			if err != nil {
+				annotations[skey], err = strconv.ParseFloat(obiutils.UnsafeString(value), 64)
+			}
+		case jsonparser.Array:
+			annotations[skey], err = _parse_json_array_interface(value)
+		case jsonparser.Object:
+			annotations[skey], err = _parse_json_map_interface(value)
+		case jsonparser.Boolean:
+			annotations[skey], err = jsonparser.ParseBoolean(value)
+		case jsonparser.Null:
+			annotations[skey] = nil
+		default:
+			log.Fatalf("Unknown data type %v", dataType)
+		}
+	}
+
+	if err != nil {
+		annotations[skey] = "NaN"
+		log.Fatalf("%s: Cannot parse value %s assicated to key %s into a %s value",
+			sequence.Id(), string(value), skey, dataType.String())
+	}
+
+	return err
+}
+
+func _parse_json_header_(header string, sequence *obiseq.BioSequence) string {
 	start := -1
 	stop := -1
 	level := 0
@@ -240,101 +343,7 @@ func _parse_json_header_(header string, sequence *obiseq.BioSequence) string {
 
 	jsonparser.ObjectEach(obiutils.UnsafeBytes(header[start:stop]),
 		func(key []byte, value []byte, dataType jsonparser.ValueType, offset int) error {
-			var err error
-
-			skey := obiutils.UnsafeString(key)
-
-			switch {
-			case skey == "id":
-				sequence.SetId(string(value))
-			case skey == "definition":
-				sequence.SetDefinition(string(value))
-
-			case skey == "count":
-				if dataType != jsonparser.Number {
-					log.Fatalf("%s: Count attribut must be numeric: %s", sequence.Id(), string(value))
-				}
-				count, err := jsonparser.ParseInt(value)
-				if err != nil {
-					log.Fatalf("%s: Cannot parse count %s", sequence.Id(), string(value))
-				}
-				sequence.SetCount(int(count))
-
-			case skey == "obiclean_weight":
-				weight, err := _parse_json_map_int(value)
-				if err != nil {
-					log.Fatalf("%s: Cannot parse obiclean weight %s", sequence.Id(), string(value))
-				}
-				annotations[skey] = weight
-
-			case skey == "obiclean_status":
-				status, err := _parse_json_map_string(value)
-				if err != nil {
-					log.Fatalf("%s: Cannot parse obiclean status %s", sequence.Id(), string(value))
-				}
-				annotations[skey] = status
-
-			case strings.HasPrefix(skey, "merged_"):
-				if dataType == jsonparser.Object {
-					data, err := _parse_json_map_int(value)
-					if err != nil {
-						log.Fatalf("%s: Cannot parse merged slot %s: %v", sequence.Id(), skey, err)
-					} else {
-						annotations[skey] = obiseq.MapAsStatsOnValues(data)
-					}
-				} else {
-					log.Fatalf("%s: Cannot parse merged slot %s", sequence.Id(), skey)
-				}
-
-			case skey == "taxid":
-				if dataType == jsonparser.Number || dataType == jsonparser.String {
-					taxid := string(value)
-					sequence.SetTaxid(taxid)
-				} else {
-					log.Fatalf("%s: Cannot parse taxid %s", sequence.Id(), string(value))
-				}
-
-			case strings.HasSuffix(skey, "_taxid"):
-				if dataType == jsonparser.Number || dataType == jsonparser.String {
-					rank := skey[:len(skey)-len("_taxid")]
-
-					taxid := string(value)
-					sequence.SetTaxid(taxid, rank)
-				} else {
-					log.Fatalf("%s: Cannot parse taxid %s", sequence.Id(), string(value))
-				}
-
-			default:
-				skey = strings.Clone(skey)
-				switch dataType {
-				case jsonparser.String:
-					annotations[skey] = string(value)
-				case jsonparser.Number:
-					// Try to parse the number as an int at first then as float if that fails.
-					annotations[skey], err = jsonparser.ParseInt(value)
-					if err != nil {
-						annotations[skey], err = strconv.ParseFloat(obiutils.UnsafeString(value), 64)
-					}
-				case jsonparser.Array:
-					annotations[skey], err = _parse_json_array_interface(value)
-				case jsonparser.Object:
-					annotations[skey], err = _parse_json_map_interface(value)
-				case jsonparser.Boolean:
-					annotations[skey], err = jsonparser.ParseBoolean(value)
-				case jsonparser.Null:
-					annotations[skey] = nil
-				default:
-					log.Fatalf("Unknown data type %v", dataType)
-				}
-			}
-
-			if err != nil {
-				annotations[skey] = "NaN"
-				log.Fatalf("%s: Cannot parse value %s assicated to key %s into a %s value",
-					sequence.Id(), string(value), skey, dataType.String())
-			}
-
-			return err
+			return _parse_json_annotation_field(key, value, dataType, sequence)
 		},
 	)
 
diff --git a/pkg/obiformats/json_read.go b/pkg/obiformats/json_read.go
new file mode 100644
index 0000000..7658ff1
--- /dev/null
+++ b/pkg/obiformats/json_read.go
@@ -0,0 +1,147 @@
+package obiformats
+
+import (
+	"io"
+	"os"
+	"path"
+
+	"git.metabarcoding.org/obitools/obitools4/obitools4/pkg/obidefault"
+	"git.metabarcoding.org/obitools/obitools4/obitools4/pkg/obiiter"
+	"git.metabarcoding.org/obitools/obitools4/obitools4/pkg/obiseq"
+	"git.metabarcoding.org/obitools/obitools4/obitools4/pkg/obiutils"
+	"github.com/buger/jsonparser"
+	"github.com/goccy/go-json"
+	log "github.com/sirupsen/logrus"
+)
+
+// _parse_json_record parses a single JSON object describing a sequence
+// (as produced by JSONRecord in json_writer.go) into a *obiseq.BioSequence.
+func _parse_json_record(raw []byte, shift byte) *obiseq.BioSequence {
+	sequence := obiseq.NewEmptyBioSequence(0)
+
+	if id, err := jsonparser.GetString(raw, "id"); err == nil {
+		sequence.SetId(id)
+	}
+
+	if seq, err := jsonparser.GetString(raw, "sequence"); err == nil {
+		sequence.SetSequence([]byte(seq))
+	}
+
+	if qual, err := jsonparser.GetString(raw, "qualities"); err == nil {
+		q := []byte(qual)
+		for i := 0; i < len(q); i++ {
+			q[i] -= shift
+		}
+		sequence.SetQualities(q)
+	}
+
+	if annot, dataType, _, err := jsonparser.Get(raw, "annotations"); err == nil && dataType == jsonparser.Object {
+		jsonparser.ObjectEach(annot,
+			func(key []byte, value []byte, valType jsonparser.ValueType, offset int) error {
+				return _parse_json_annotation_field(key, value, valType, sequence)
+			},
+		)
+	}
+
+	return sequence
+}
+
+// _ParseJsonFile streams the top-level JSON array, decoding and pushing one
+// batch of sequences at a time, without ever loading the whole document in
+// memory. Only one raw record at a time is buffered by the decoder.
+func _ParseJsonFile(source string,
+	reader io.Reader,
+	out obiiter.IBioSequence,
+	shift byte,
+	batchSize int) {
+
+	dec := json.NewDecoder(reader)
+
+	if _, err := dec.Token(); err != nil {
+		if err == io.EOF {
+			out.Done()
+			return
+		}
+		log.Fatalf("cannot parse JSON data: %v", err)
+	}
+
+	slice := obiseq.MakeBioSequenceSlice()
+	o := 0
+
+	for dec.More() {
+		var raw json.RawMessage
+
+		if err := dec.Decode(&raw); err != nil {
+			log.Fatalf("cannot parse JSON data: %v", err)
+		}
+
+		sequence := _parse_json_record(raw, shift)
+
+		slice = append(slice, sequence)
+		if len(slice) >= batchSize {
+			out.Push(obiiter.MakeBioSequenceBatch(source, o, slice))
+			o++
+			slice = obiseq.MakeBioSequenceSlice()
+		}
+	}
+
+	if len(slice) > 0 {
+		out.Push(obiiter.MakeBioSequenceBatch(source, o, slice))
+	}
+
+	out.Done()
+}
+
+func ReadJSON(reader io.Reader, options ...WithOption) (obiiter.IBioSequence, error) {
+
+	opt := MakeOptions(options)
+	out := obiiter.MakeIBioSequence()
+
+	out.Add(1)
+	go _ParseJsonFile(opt.Source(),
+		reader,
+		out,
+		obidefault.ReadQualitiesShift(),
+		opt.BatchSize())
+
+	go func() {
+		out.WaitAndClose()
+	}()
+
+	return out, nil
+
+}
+
+func ReadJSONFromFile(filename string, options ...WithOption) (obiiter.IBioSequence, error) {
+
+	options = append(options, OptionsSource(obiutils.RemoveAllExt((path.Base(filename)))))
+	file, err := obiutils.Ropen(filename)
+
+	if err == obiutils.ErrNoContent {
+		log.Infof("file %s is empty", filename)
+		return ReadEmptyFile(options...)
+	}
+
+	if err != nil {
+		return obiiter.NilIBioSequence, err
+	}
+
+	return ReadJSON(file, options...)
+}
+
+func ReadJSONFromStdin(reader io.Reader, options ...WithOption) (obiiter.IBioSequence, error) {
+	options = append(options, OptionsSource(obiutils.RemoveAllExt("stdin")))
+	input, err := obiutils.Buf(os.Stdin)
+
+	if err == obiutils.ErrNoContent {
+		log.Infof("stdin is empty")
+		return ReadEmptyFile(options...)
+	}
+
+	if err != nil {
+		log.Fatalf("open file error: %v", err)
+		return obiiter.NilIBioSequence, err
+	}
+
+	return ReadJSON(input, options...)
+}
diff --git a/pkg/obiformats/universal_read.go b/pkg/obiformats/universal_read.go
index 62e967e..7df141c 100644
--- a/pkg/obiformats/universal_read.go
+++ b/pkg/obiformats/universal_read.go
@@ -145,6 +145,8 @@ func ReadSequencesFromFile(filename string,
 		return ReadGenbank(reader, options...)
 	case "text/csv":
 		return ReadCSV(reader, options...)
+	case "application/json":
+		return ReadJSON(reader, options...)
 	default:
 		log.Fatalf("File %s has guessed format %s which is not yet implemented",
 			filename, mime.String())
diff --git a/pkg/obitools/obiconvert/options.go b/pkg/obitools/obiconvert/options.go
index a860194..a7fbe02 100644
--- a/pkg/obitools/obiconvert/options.go
+++ b/pkg/obitools/obiconvert/options.go
@@ -24,6 +24,7 @@ var __input_genbank_format__ = false
 var __input_fastq_format__ = false
 var __input_fasta_format__ = false
 var __input_csv_format__ = false
+var __input_json_format__ = false
 
 var __output_in_fasta__ = false
 var __output_in_fastq__ = false
@@ -71,6 +72,9 @@ func InputOptionSet(options *getoptions.GetOpt) {
 	options.BoolVar(&__input_csv_format__, "csv", __input_csv_format__,
 		options.Description("Read data following the CSV format."))
 
+	options.BoolVar(&__input_json_format__, "json", __input_json_format__,
+		options.Description("Read data following the JSON format."))
+
 	options.BoolVar(&__no_ordered_input__, "no-order", __no_ordered_input__,
 		options.Description("When several input files are provided, "+
 			"indicates that there is no order among them."))
@@ -158,6 +162,8 @@ func CLIInputFormat() string {
 		return "genbank"
 	case __input_csv_format__:
 		return "csv"
+	case __input_json_format__:
+		return "json"
 	default:
 		return "guessed"
 	}
diff --git a/pkg/obitools/obiconvert/sequence_reader.go b/pkg/obitools/obiconvert/sequence_reader.go
index cbda802..d16f39e 100644
--- a/pkg/obitools/obiconvert/sequence_reader.go
+++ b/pkg/obitools/obiconvert/sequence_reader.go
@@ -73,7 +73,9 @@ func ExpandListOfFiles(check_ext bool, filenames ...string) ([]string, error) {
 						strings.HasSuffix(path, "dat") ||
 						strings.HasSuffix(path, "dat.gz") ||
 						strings.HasSuffix(path, "ecopcr") ||
-						strings.HasSuffix(path, "ecopcr.gz") {
+						strings.HasSuffix(path, "ecopcr.gz") ||
+						strings.HasSuffix(path, "json") ||
+						strings.HasSuffix(path, "json.gz") {
 						log.Debugf("Appending %s file\n", path)
 						list_of_files.Add(path)
 					}
@@ -142,6 +144,8 @@ func CLIReadBioSequences(filenames ...string) (obiiter.IBioSequence, error) {
 			iterator, err = obiformats.ReadFastq(os.Stdin, opts...)
 		case "csv":
 			iterator, err = obiformats.ReadCSV(os.Stdin, opts...)
+		case "json":
+			iterator, err = obiformats.ReadJSON(os.Stdin, opts...)
 		default:
 			iterator, err = obiformats.ReadSequencesFromStdin(opts...)
 		}
@@ -163,6 +167,8 @@ func CLIReadBioSequences(filenames ...string) (obiiter.IBioSequence, error) {
 			reader = obiformats.ReadFastaFromFile
 		case "csv":
 			reader = obiformats.ReadCSVFromFile
+		case "json":
+			reader = obiformats.ReadJSONFromFile
 		case "ecopcr":
 			reader = obiformats.ReadEcoPCRFromFile
 		case "embl":
diff --git a/pkg/obiutils/mimetypes.go b/pkg/obiutils/mimetypes.go
index 52cf083..901dea2 100644
--- a/pkg/obiutils/mimetypes.go
+++ b/pkg/obiutils/mimetypes.go
@@ -102,6 +102,11 @@ func RegisterOBIMimeType() {
 			return ok
 		}
 
+		jsonDetector := func(raw []byte, limit uint32) bool {
+			raw = bytes.TrimLeft(raw, " \t\r\n")
+			return len(raw) > 0 && (raw[0] == '[' || raw[0] == '{')
+		}
+
 		mimetype.Lookup("text/plain").Extend(fastaDetector, "text/fasta", ".fasta")
 		mimetype.Lookup("text/plain").Extend(fastqDetector, "text/fastq", ".fastq")
 		mimetype.Lookup("text/plain").Extend(ecoPCR2Detector, "text/ecopcr2", ".ecopcr")
@@ -115,6 +120,7 @@ func RegisterOBIMimeType() {
 		mimetype.Lookup("application/octet-stream").Extend(genbankDetector, "text/genbank", ".seq")
 		mimetype.Lookup("application/octet-stream").Extend(emblDetector, "text/embl", ".dat")
 		mimetype.Lookup("application/octet-stream").Extend(csv, "text/csv", ".csv")
+		mimetype.Lookup("application/octet-stream").Extend(jsonDetector, "application/json", ".json")
 	}
 	__obimimetype_registred__ = true
 }