Bringing AI-Powered Voice Conversations to Mobile Video Advertising
Diffco built the iOS and Android delivery layer for Instreamatic's interactive video ads — reusable SDK components and adaptive HLS streaming that let an AI-driven, two-way ad conversation start instantly, stay responsive, and keep playing across devices and network conditions.
Project summary
Instreamatic pioneered "dialogue advertising" — ads a listener can talk back to.
Its voice AI interprets an open-ended spoken reply, detects intent and tone, and plays the most relevant next message, turning a passive impression into a measurable conversation.
Having proven the format in audio, Instreamatic set out to bring it to video inside mobile apps. Diffco partnered with the Instreamatic team to engineer the production-ready mobile experience: new iOS and Android SDK components, an HLS-based adaptive streaming architecture, and the real-time interaction loop that ties playback, microphone capture, and the AI response into one continuous experience.
Client summary
About Instreamatic
Founded in 2015 and headquartered in the Bay Area before moving to Boca Raton, Florida, Instreamatic built a voice-AI platform that lets brands hold two-way conversations with consumers inside audio and video streams. Its later product line used generative AI to turn a single ad recording into hundreds of personalized, contextual variations for audio, video, and connected TV.
The company's platform was backed by Progress Ventures, Accomplice, and Google Assistant Investments, ran campaigns for brands including IKEA, Infiniti, HP, Intel, Mercedes, and NetSuite, and distributed through publishers such as Pandora, Salem Media Group, Gaana, Triton Digital, and DAX.
- 12%Average voice engagement rate on interactive ads
- 19%Peak campaign engagement rate (Infiniti)
- 0.6%Industry benchmark: average mobile ad click-through rate
- +22 pptBrand favorability lift from personalized AI-generated ads (Veritonic study with dentsu and Intel)
- +18 pptPurchase intent lift from personalized AI-generated ads
- 3 min vs. 4–6 weeksAd production time on the platform vs. traditional studio production
Top achievements
Pandora partnership (2019)
Agreement with the largest streaming service in the U.S. to test AI-powered interactive voice ads, with Doritos, Ashley HomeStore, and Hellmann's among the brands running dialogue campaigns.
Salem Media Group rollout (2019)
Voice dialogue ads introduced across Salem's 115 owned stations and roughly 2,700 syndicated affiliates.
World's first programmatically executed dialogue ad campaign
Delivered with media agency Essence on Gaana, India's most-used music streaming service; a Nielsen ad-effectiveness study showed significant uplift in brand consideration and intent to use.
AdExchanger Awards finalist (2021)
for Innovation in Audio Advertising.
Client goals
Client Goals for Interactive Video Advertising
Extend a proven audio format to video
Take interactive voice advertising into mobile video without losing the immediacy that made the audio product work.
Ship production-grade native SDKs
Give app publishers drop-in iOS and Android components so integration takes days, not months.
Deliver reliable video under real-world networks
Start instantly, stream without freezes, and degrade gracefully as bandwidth changes.
Protect the conversation
Capture a voice response and branch to the next creative without interrupting content playback.
Keep every impression measurable
Surface each interaction as a signal advertisers can act on.
The challenge
Engineering the Seamless Interactive Video Experience
The intelligence behind an interactive ad is only part of the product. For the format to work on a phone, the ad has to start the moment the slot opens, stream a video reliably, open the microphone at the right moment, hand the spoken reply to the AI, and play the branching response — all without a visible seam, on thousands of device and network combinations.
Video also raised the stakes. AI-driven branching means the creative is dynamic, not a single pre-buffered file, so playback reliability and adaptive quality became first-order engineering problems rather than polish.
Our contribution
Building the Production-Ready Interactive Video Stack
Native SDK components for iOS and Android
Together with the Instreamatic team, Diffco developed new components for the voice and video advertising SDKs. They package playback, microphone access, user input, and the platform response behind a single integration surface, so publishers can deploy interactive ads into their apps faster and with interoperability across brands and products.
Adaptive HLS streaming architecture
Diffco implemented HLS-based streaming with adaptive bitrate support, so video quality adjusts automatically to available bandwidth. The system tunes compression and resolution on the fly to maintain the best viewing experience the connection allows.
The real-time interaction loop
Ad starts → consumer responds → voice is processed → intent is detected → relevant response plays → interaction is measured. Diffco connected the media playback and interaction layers so this loop feels like one continuous experience, with the SDK coordinating streaming, audio capture, and the AI response while protecting playback continuity.
Playback resilience
Handling for orientation changes, subtitles, and continuity across state transitions preserves the conversational experience without freezes or unexpected quality drops.
AI-Driven Interaction
How AI powers the experience
Natural language understanding
Voice AI processes open-ended consumer responses instead of relying on predefined buttons or commands.
Intent and tone detection
The platform identifies what the consumer means and how they respond, then continues along the most relevant conversational path.
Contextual creative generation
Generative AI produces multiple variations of an audio or video ad, adapting messaging to the content, environment, location, platform, or campaign conditions.
Continuous optimization
Every interaction creates a measurable signal that helps advertisers understand engagement and improve future campaigns.
Benefits for the business
Business Impact of the Video Expansion
A new product line on an existing platform
Interactive video ads gave Instreamatic a second format to sell to the brands and publishers already running its voice audio campaigns.
Faster publisher onboarding
Reusable SDK components turned each new app integration into a repeatable deployment instead of a bespoke project.
Consistent quality across the long tail of devices
Adaptive streaming kept the experience intact for users on weak or fluctuating connections, where interactive formats are most fragile.
Results
What the Partnership Delivered
Interactive voice and video ad components shipped for both iOS and Android SDKs and made available to app publishers.
HLS adaptive streaming delivered stable playback and automatic quality adjustment across a wide range of devices and connection speeds.
The end-to-end interaction loop — playback, voice capture, AI response, measurement — ran as a single continuous experience with playback continuity preserved through orientation changes and network shifts.
Instreamatic's platform went on to reach a 12% average voice engagement rate (19% at peak) and earned an AdExchanger Awards finalist spot for audio innovation.
Technologies used
- Objective-C
- C/C++
- Java
- iOS SDK
- Android SDK
- HLS video streaming protocol
They were communicative, consultative, and helpful. They went above and beyond all the way. What an amazing team. I will work with them again and again.
Stas TushinskiyCEO & Co-founder, Instreamatic
