diff --git a/bcbio/bam/fastq.py b/bcbio/bam/fastq.py
index ce7545d5c..b6261113f 100644
--- a/bcbio/bam/fastq.py
+++ b/bcbio/bam/fastq.py
@@ -220,6 +220,9 @@ def estimate_read_length(fastq_file, quality_format="fastq-sanger", nreads=1000)
     read = in_handle.next()
     average = len(read.seq)
     for _ in range(nreads):
-        average = (average + len(in_handle.next().seq)) / 2
+        try:
+            average = (average + len(in_handle.next().seq)) / 2
+        except StopIteration:
+            break
     in_handle.close()
     return average
diff --git a/bcbio/ngsalign/star.py b/bcbio/ngsalign/star.py
index d813ee84b..6c847a3ba 100644
--- a/bcbio/ngsalign/star.py
+++ b/bcbio/ngsalign/star.py
@@ -1,21 +1,19 @@
 from os import path
-import subprocess
+import tempfile
 
 from bcbio.pipeline import config_utils
-from bcbio.utils import safe_makedir, file_exists
+from bcbio.utils import safe_makedir, file_exists, get_in
 from bcbio.provenance import do
+from bcbio.bam import fastq
 
 CLEANUP_FILES = ["Aligned.out.sam", "Log.out", "Log.progress.out"]
 
-fastq_file = "/Users/rory/tmp/data/test_fastq_1.fastq"
-pair_file = "/Users/rory/tmp/data/test_fastq_2.fastq"
-align_dir = "/Users/rory/tmp/star_test"
-ref_file = "/Users/rory/tmp/GRCm38/STAR"
-out_base = "/Users/rory/tmp/star_test/test"
 
 
 def align(fastq_file, pair_file, ref_file, out_base, align_dir, data,
           names=None):
+    genome_dir = create_genome(fastq_file, data)
+
     config = data["config"]
     out_prefix = path.join(align_dir, out_base)
     out_file = out_prefix + "Aligned.out.sam"
@@ -25,15 +23,50 @@ def align(fastq_file, pair_file, ref_file, out_base, align_dir, data,
     fastq = " ".join([fastq_file, pair_file])
     num_cores = config["algorithm"].get("num_cores", 1)
     safe_makedir(align_dir)
-    cmd = ("{star_path} --genomeDir {ref_file} --readFilesIn {fastq} "
+    cmd = ("{star_path} --genomeDir {genome_dir} --readFilesIn {fastq} "
            "--runThreadN {num_cores} --outFileNamePrefix {out_prefix} "
            "--outReadsUnmapped Fastx --outFilterMultimapNmax 10")
     run_message = "Running STAR aligner on %s and %s." % (pair_file, ref_file)
     do.run(cmd.format(**locals()), run_message, None)
     return out_file
 
+def _get_quality_format(config):
+    qual_format = config["algorithm"].get("quality_format", None)
+    if qual_format.lower() == "illumina":
+        return "fastq-illumina"
+    elif qual_format.lower() == "solexa":
+        return "fastq-solexa"
+    else:
+        return "fastq-sanger"
 
 def remap_index_fn(ref_file):
     """Map sequence references to equivalent star indexes
     """
     return path.join(path.dirname(path.dirname(ref_file)), "star")
+
+def create_genome(fastq_file, data):
+    """
+    create a genome file with splicing information. this depends on read length
+    so must be made custom for each mapping
+    """
+    config = data["config"]
+    star_path = config_utils.get_program("STAR", config)
+    tempdir = tempfile.mkdtemp("_star")
+    quality_format = _get_quality_format(data["config"])
+    overhang = fastq.estimate_read_length(fastq_file, quality_format, 100000) - 1
+    fasta_file = data["sam_ref"]
+    gtf_file = get_in(data, ("genome_resources", "rnaseq", "transcripts"), None)
+    num_cores = config["algorithm"].get("num_cores", 1)
+
+    mem_limit = 1000000000
+
+    cmd = ("{star_path} --genomeDir {tempdir} --genomeFastaFiles {fasta_file} "
+           " --runThreadN {num_cores} --runMode genomeGenerate "
+           "--sjdbOverhang {overhang} --limitGenomeGenerateRAM {mem_limit}")
+    if gtf_file:
+        cmd + " --sjdbGTFfile {gtf_file}"
+    run_message = ("Generating a STAR index for %s using %s as transcripts with an "
+                   "overhang of %d") % (fasta_file, gtf_file, overhang)
+
+    do.run(cmd.format(**locals()), run_message, None)
+    return tempdir