A clear podcast doesn’t start with expensive software. It starts with a clean recording, controlled processing, and the right export settings.
If voices sound distant, harsh, boomy, or uneven, listeners notice within seconds. You can fix most spoken-word problems before uploading to Transistor.fm with a short, repeatable audio workflow.
Start with the source file, apply each processor in the correct order, then check the finished episode on more than one device.
Key Takeaways
- Transistor.fm supports common audio uploads, but MP3 is the safest delivery format for most podcasts.
- Clean recording technique reduces the amount of noise reduction and processing you need.
- Use EQ, compression, de-essing, and loudness normalization in small steps.
- Export a final listening copy, but retain the original WAV files and project sessions.
- Check the finished episode before publishing it through your Transistor RSS feed.
Check Transistor.fm’s Audio Requirements First
Transistor.fm accepts common podcast audio formats, including MP3, M4A, WAV, and FLAC. Its current documentation lists a maximum audio upload size of 2 GB. You should still check the Transistor support center before a major production change because upload rules can change.
For normal spoken-word episodes, MP3 is the practical choice. Export at 44.1 kHz with a constant bitrate. A 128 kbps mono file works well for a single-voice show. Use 128 to 192 kbps stereo when your episode includes music, multiple channels, or a wider sound design.
Transistor doesn’t require you to upload a finished MP3 at a specific loudness target. The responsibility stays with your production workflow. Uploading an unprocessed WAV won’t automatically fix clipping, room noise, sharp S sounds, or inconsistent volume.
Apple’s official podcast audio requirements are also useful when you distribute through an RSS feed. They cover accepted formats, sample rates, bitrates, and other delivery details.
Keep these files separate:
- Your original recording, preferably WAV or another lossless format.
- Your edited master, saved before final encoding.
- Your delivery file, such as the MP3 uploaded to Transistor.
The delivery file is replaceable. The original recording is not. Retain backups in two locations, such as a local drive and cloud storage. Keep the project session as well if you may need to repair a sentence, remove a cough, or create a trailer later.
Record Clean Speech Before You Process It
Audio software can’t fully repair a clipped recording. It also can’t remove room reflections without affecting the voice. Your first improvement should happen before you press record.
Place the microphone 10 to 15 centimeters from your mouth. Point it slightly to the side instead of directly at your lips. This reduces plosive sounds from words that begin with P and B. Use a pop filter when possible.
Keep the microphone and your mouth at a consistent distance. Moving closer changes the bass response. Moving away makes the room louder and the voice thinner.
Record in a room with soft furnishings. Curtains, carpets, sofas, and bookshelves reduce reflections. Bare walls and desks create the hollow sound that noise reduction often fails to remove.
Set the input gain so normal speech peaks around -12 to -6 dBFS. Leave headroom for loud words and sudden laughter. Red meters or flattened waveforms indicate clipping. Reduce the input gain and record again.
A quiet recording with moderate peaks gives you more control than a loud recording with distortion.
Before the full session, record 20 seconds of room tone. Room tone is the sound of the space when nobody speaks. Noise reduction tools use this sample to identify steady background noise.
Use headphones during recording. Listen for computer fans, chair movement, cable noise, mouth clicks, and electrical hum. Fix the source before you start editing.
Enhance Speech Audio With a Controlled Processing Chain
Process the voice in a consistent order. Make a copy of the edited file first, then apply each change while watching both the meters and the waveform.
1. Reduce steady background noise
Start with a noise reduction tool in Audacity, Adobe Audition, Hindenburg, Descript, or another editor. Select the room-tone section and capture a noise profile if the software uses one.
Apply a light reduction first. A range of 4 to 8 dB is a useful starting point for steady fan noise or air conditioning. Larger reductions can create metallic artifacts, watery tails, and unnatural gaps between words.
Avoid trying to remove every trace of room sound. A small amount of consistent background noise is less distracting than a voice damaged by aggressive processing.
If the noise changes throughout the recording, split the episode into sections and process them separately. A single profile may not work for a fan, traffic, and handling noise at the same time.
2. Use EQ to remove problems before adding tone
Equalization changes the balance of frequencies. It shouldn’t turn a weak recording into a different voice.
Apply a high-pass filter between 70 and 100 Hz. This removes low rumble from footsteps, desks, traffic, and microphone handling. Don’t set the filter too high or the voice may lose body.
If the speech sounds muddy, reduce 2 to 4 dB around 200 to 350 Hz. If it sounds nasal, check the 700 Hz to 1.2 kHz range with a small cut. Use a narrow adjustment only when you can hear a clear problem.
For more presence, try a gentle 1 to 3 dB boost between 2 and 4 kHz. Stop when consonants become clear. A large boost creates listener fatigue.
Every voice and microphone combination is different. Bypass the EQ often. If the processed version only sounds louder, not clearer, undo the change.
3. Add compression for consistent volume
Compression reduces the difference between quiet and loud speech. It helps listeners hear a steady voice without constantly adjusting their device volume.
Start with a ratio between 2:1 and 4:1. Set the threshold so normal speech receives around 3 to 6 dB of gain reduction. Use an attack of roughly 10 to 30 milliseconds and a release between 80 and 150 milliseconds.
These are starting points, not fixed rules. A quiet interview may need more control than a tightly delivered solo show. Listen for pumping, audible breaths, and a flattened delivery.
Add makeup gain only after the compressor behaves correctly. Increasing the output too early can make you mistake loudness for improvement.
4. Apply de-essing after compression
Compression can make S, SH, and CH sounds more noticeable. A de-esser reduces those sharp frequencies without lowering the whole voice.
Set the detection range around 4.5 to 8 kHz, then target only the harsh consonants. Aim for about 2 to 5 dB of reduction on strong S sounds.
If the voice starts sounding dull or lispy, reduce the intensity. De-essing should control sibilance, not erase normal speech detail.
5. Normalize loudness and limit peaks
Set a loudness target after the main processing is complete. Many podcast workflows use -16 LUFS integrated for stereo audio and -19 LUFS for mono audio. Pick one standard for your show and use it consistently.
Use a true-peak limiter at the end of the chain. Set the ceiling around -1 dBTP. This helps prevent peaks created during encoding from clipping on playback.
Measure the full episode, not a single sentence. Music, silence, interviews, and ad breaks all affect the integrated loudness reading.
You can use a loudness meter in your editor or an automated service such as Auphonic’s audio processing platform. Automated processing can save time, but you still need to listen to the result. No loudness meter can tell you that a voice sounds metallic or that a breath was cut unnaturally.
Export, Test, and Upload the Episode
Finish with a lossless archive and a separate delivery file. Export the edited master as WAV when your editor supports it. Then create the MP3 for Transistor.fm.
Use a constant bitrate instead of a variable bitrate when you want predictable file behavior. Check the MP3 properties before uploading. Confirm the sample rate, channel format, bitrate, and file duration.
Listen to the export, not only the project session. Encoding can expose clicks, clipped transitions, or missing audio that wasn’t obvious during editing.
Use this final review:
- Listen through headphones for clicks, harsh S sounds, edits, and background changes.
- Listen through a phone or laptop speaker for voice clarity.
- Check the first and last 30 seconds for accidental silence or abrupt cuts.
- Confirm that the voice remains clear under music and advertisements.
- Read the loudness and true-peak meters from the complete file.
- Compare the episode with one of your recent published episodes.
Transistor’s podcast hosting service lets you upload the finished episode to the correct show, add the title and description, then publish or schedule it. Review the episode details before confirming publication. A technically clean file can still create a poor listener experience if it carries the wrong show information or release date.
Don’t overwrite your source files after upload. Store the original recording, edited master, and final MP3 with clear filenames. For example:
show-042-original.wavshow-042-edited-master.wavshow-042-transistor.mp3
This naming system makes future corrections easier. It also prevents you from processing an already compressed MP3 a second time.
Avoid the Processing Mistakes That Damage Speech
The most common mistake is using every effect at maximum strength. Noise reduction removes too much room tone. EQ boosts too much treble. Compression removes natural dynamics. De-essing makes the voice dull.
Process one problem at a time. Bypass each effect and compare it with the untreated version. If you can’t hear a clear improvement, remove the effect.
Don’t use a noise gate to hide every pause. A gate can cut word endings and make room tone jump in and out. Manual editing or a gentle expander usually produces a more natural result.
Don’t normalize peaks to 0 dBFS. Peak normalization doesn’t create consistent perceived volume. Use loudness measurement and leave headroom for encoding.
One-click speech enhancers can help with difficult recordings, but use them on a copy. They often change room tone, breaths, and vocal texture. A clean original gives you a way back when the result sounds artificial.
Conclusion
You can enhance speech audio for Transistor.fm without a complex studio setup. Record with headroom, reduce noise lightly, shape the voice with EQ, control dynamics with compression, soften sibilance, and finish with a consistent loudness target.
Export a tested MP3 for publishing, but keep the original WAV files and project sessions backed up. The best processing chain is the one that makes the speaker easier to understand without making listeners notice the processing.
