Introducing Tonn SDK v2: AI-Powered Audio Post-Production for Film, TV, and Advertising

Since 2023, the Tonn platform has processed over five million music tracks through our mixing and mastering engine. Today, we're launching Tonn SDK v2, a purpose-built post-production module that brings that same audio intelligence to film, television, streaming, advertising, and podcasting workflows.

This isn't a repurposed music tool. It's an entirely new processing pipeline designed around the way post-production actually works: dialogue-first, delivery-spec compliant, and built for the formats that broadcasters and platforms demand.

The post-production module is available now in the Tonn C++ SDK, with cloud API availability coming in the next few weeks.

Why post-production needs a new approach

Audio post-production has a consistency problem. Every project has its own mix of dialogue, music, sound effects, foley, and ambience, and every delivery platform has its own loudness spec. Getting a mix to sound right and pass QC for Netflix, YouTube, broadcast TV, and cinema means hours of manual work per deliverable.

We built Tonn SDK v2 to make that process programmable. One API call. Correct loudness. Compliant delivery. Processed stems if you need them.

What's in the box

Native post-production track types

The SDK speaks the language of post-production natively, with 10 dedicated track categories: Dialogue (main and secondary), Music, SFX, Ambience, Foley, ADR, Voiceover (low-pitch and high-pitch variants), and a general-purpose category. This means the processing engine understands what it's working with and treats dialogue differently from a music bed or a foley track.

Intelligent dialogue enhancement

Three processing modes tuned to different content types:

• Preserve for documentaries and naturalistic content, where minimal intervention keeps the original character intact

• Enhance for standard TV, streaming, and interview content, applying a clarity boost without overprocessing

• AD Enhanced for commercials, announcements, and voiceover, pushing intelligibility as far as it can go

Under the hood, the engine performs noise floor detection, frequency balance assessment, clipping analysis, and reverb profiling before applying any processing. It adapts to what it hears.

ADR-to-location sound matching

One of the hardest tasks in post-production is making re-recorded dialogue (ADR) sit naturally against original location sound. The SDK's DialogueMatcher analyses frequency response, dynamics, timbre, and reverb characteristics of the original recording, then applies matching processing to the ADR take. It reports similarity scores back so you can verify the match programmatically.

Automatic sidechain ducking

Background audio (music, SFX, ambience) automatically ducks when dialogue is detected. Five presets cover common scenarios, from a subtle -6 dB reduction for background music through to an aggressive -20 dB cut for narration over score. A fully custom mode gives you control over reduction depth, attack, and release. Multi-track ducking is supported, so one dialogue sidechain can control multiple background tracks simultaneously.

Seven delivery format presets

Each preset is calibrated to a specific delivery standard with loudness targets, true peak limits, and LRA (Loudness Range) constraints:

The output includes a pass/fail compliance report covering programme loudness, dialogue loudness, true peak, programme LRA, and dialogue LRA. You can surface this directly to end users or feed it into your QC pipeline.

Processed stems output

Optionally export individually processed stems alongside the final mix. Useful for QC, revision, and downstream editorial work.

GPU acceleration

Processing is GPU-accelerated on both macOS (Metal) and Linux (CUDA), with automatic fallback to CPU when no GPU is available.

Hear it in action

Here are two before-and-after examples showing the post-production module at work.

Example 1: Advertising spot

This example uses the Ad Mix delivery profile with sidechain ducking enabled, using the dialogue track as the sidechain source. The mastering stage targets EBU R128 for advertising.

Example 2: Animation for streaming

This example uses the streaming delivery profile with no ducking applied. The mastering targets the major streaming platform's loudness spec.

Available now in the C++ SDK. Cloud API coming soon.

The post-production module ships today in the Tonn C++ SDK with full documentation, example applications, and JSON configuration templates for common workflows.

We're bringing these capabilities to our cloud API in the coming weeks. Post-production services will use the same credit system as our existing music API, so any customer with API credits can start using the new endpoints immediately. No separate plan or licence required.

What's next

We have a clear roadmap for extending the post-production module:

• Cloud API launch in the coming weeks, starting with dialogue enhancement, delivery format mastering, and loudness compliance

• Full pipeline endpoints, including sidechain ducking, ADR matching, and the bundled post-production mix, shortly after

• Surround and immersive audio support is on our roadmap (the delivery format metadata already carries surround and channel count fields)

If you're building tools for post-production, broadcast, podcasting, or video platforms, we'd love to talk. The SDK documentation is live, and you can register for API access at tonn-portal.roexaudio.com.

The Tonn SDK and API are built by RoEx, the audio technology company behind Automix and Mix Check Studio. We've processed over 5 million tracks since 2023.