🥪 Merge Audio

You generated a narration, you generated some ambience, and now you need them in one track. Merge Audio is the little mixing-desk node that does it - combine up to four audio sources into one, with a choice of how they blend. Layer a TTS voice over background music, stack a sound effect under dialogue, or crossfade between two clips. It's not a full DAW, but for finishing a ComfyUI audio graph without exporting to something else, it's exactly enough.

It handles the annoying plumbing too: mismatched sample rates get reconciled automatically, and it normalizes the result so you don't blow past 0 dB and clip.

How it works

You feed it audio and pick a merge_algorithm - the math it uses to combine the sources. The two you'll reach for most are overlay (natural layering - voice on top of music, both audible) and crossfade (smooth transition from one clip into the next). The others (mean, median, max, min, sum, weighted) are for specific jobs: mean for a balanced average, median to knock down outlier noise, weighted when you want to control the balance yourself. For most "voice + music" work, overlay is the answer.

The inputs and outputs that matter

• audio1 and audio2 - required, the two sources. audio3 and audio4 are optional for stacking more.
• merge_algorithm - the blend mode. overlay for layering, crossfade for transitions, weighted for a manual mix. This is the main decision.
• volume_balance (default 0.5) - the balance between sources. Slide it toward one input to make it louder in the mix - the lever for "voice too quiet under the music."
• normalize (on) - keeps the merged output from clipping. Leave it on unless you have a reason not to.

For transitions, crossfade_duration sets how long the blend lasts. There's also an optional pitch section - enable_pitch_control plus vocal_pitch_shift / instrumental_pitch_shift - for shifting a source up or down in semitones, and sample_rate (default auto) to force an output rate if you need a specific one.

Two outputs: merged_audio (AUDIO) → your Save/Preview Audio node, and merge_info (text) describing what it did.

Installing it

Comes with the pack. ComfyUI Manager → search "TTS Audio Suite" → install → restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/diodiogod/TTS-Audio-Suite.git
cd TTS-Audio-Suite
python install.py

No model download - it's plain audio DSP, so it's instant once the pack is installed.

Common issues & troubleshooting

The voice is buried under the music. That's volume_balance. Slide it toward the voice input so it sits on top. And make sure you're on overlay, not an averaging mode like mean that pulls everything toward a middle level.

The result sounds crushed or pumping. Normalization is preventing clipping, but if two loud sources sum hot, the normalize pass can squash things. Balance the inputs first (with volume_balance or by trimming a source upstream) rather than relying on normalize to fix a too-hot mix.

A hard cut where I wanted a smooth blend. Use crossfade as the algorithm and give crossfade_duration enough time (it defaults short) so the transition actually fades rather than jumping.

Sample-rate weirdness. It auto-reconciles rates by default, which is usually right. Only set sample_rate manually if a downstream node demands a specific rate - otherwise leave it on auto and let it match.