Google Supercharges AI for Developers: Gemini Omni 1.1 Flash for Video & 3.5 Transcribe for Speech Hit GA

The General Availability of Gemini Omni 1.1 Flash, accessible via Google AI Studio and the Gemini API, is a game-changer for developers looking to integrate advanced video generation and editing into their applications. This release isn't just about generating video; it's about providing unprecedented control over the creative process, moving AI video closer to a traditional editing workflow.

Key capabilities introduced in Omni 1.1 Flash include:

• Video Extension: Developers can now seamlessly extend existing video clips in 10-second increments, allowing for longer narratives up to a cumulative 40 seconds. This feature boasts improved visual consistency and narrative adherence by analyzing up to 10 seconds of prior context, a significant leap from previous models that often only considered the final second.
• First and Last Frame Interpolation: This powerful feature enables developers to provide a starting image and an ending image, with the model generating a smooth, continuous video transition between them. This is ideal for creating complex camera movements, zoom transitions, or seamless looping clips, offering a level of control akin to keyframe animation.
• Resolution Control & Drafts: The model supports selectable output resolutions including 360p, 720p (default), 1080p, and 4K. Crucially, 1080p and 4K outputs are generated via upscaling, allowing developers to create quick, cost-effective 360p drafts for rapid iteration (at approximately one-third the cost of 720p) before committing to higher-resolution final renders.
• Video References: Users can provide up to three seconds of reference video as input, which helps the model maintain consistent motion, characters, and overall visual context throughout the generated clip.
• Conversational Editing: Enabled by the Interactions API, this capability allows for iterative refinement and editing of videos through natural language conversations. Developers can describe desired changes, and the model applies the edits while preserving other parts of the video.

The pricing model for Omni 1.1 Flash is token-based, with estimated per-second costs varying by resolution: $0.03 at 360p, $0.10 at 720p, $0.15 at 1080p, and $0.30 at 4K. This transparent pricing, combined with the draft-then-upscale workflow, is designed to help teams budget and integrate AI video generation more predictably into their pipelines.

Alongside the video advancements, Google also announced the General Availability of Gemini 3.5 Transcribe, a new suite of speech-to-text models that significantly raises the bar for accuracy and functionality. Designed for intelligent voice interactions, this model excels in challenging real-world environments, handling background noise, complex jargon, and natural speech disfluencies with remarkable precision.

Gemini 3.5 Transcribe is offered through two distinct API endpoints, catering to different use cases:

• gemini-3.5-transcribe (Interactions API): This endpoint is optimized for high-accuracy, low-latency non-streaming speech-to-text, ideal for processing pre-recorded audio files. It supports utterance-based language detection across over 85 languages, speaker diarization for up to three (and experimentally more) speakers, word-level timestamps, and custom vocabulary biasing (up to 1,000 terms).
• gemini-3.5-transcribe-live (Live API): Tailored for real-time voice interfaces, this endpoint provides low-latency, bidirectional streaming speech-to-text over WebSockets. It supports interim and finalized transcription events, a 'Smart transcription' mode that cleans up filler words and auto-formats text, and multiple Voice Activity Detection (VAD) strategies.

The model boasts impressive performance metrics, with an average Word Error Rate (WER) of 2.6% for non-streaming audio and 4.0% for streaming use cases, as measured by Artificial Analysis. Google reports a 70% improvement in time to final transcription compared to its previous Chirp 3 model. The estimated pricing is developer-friendly, with recorded audio transcription at approximately $0.005 per minute and live transcription at $0.009 per minute, with a free tier available for solo developers and startups.

These releases from Google highlight a strategic focus on empowering developers with more granular control and practical tools for integrating advanced AI into their applications. The emphasis on iterative workflows in Gemini Omni 1.1 Flash, with features like scene extension and frame interpolation, reflects a maturation in generative AI for media. Developers can now approach AI video creation not as a black box, but as a malleable process where they can guide and refine the output, significantly reducing the 'reroll and hope' cycles that characterized earlier models. This shift is crucial for professional content creation, game development, marketing, and any domain requiring precise visual storytelling.

Similarly, Gemini 3.5 Transcribe's high accuracy, multi-language support, and features like speaker diarization and custom vocabulary biasing make it invaluable for a wide array of applications. From enhancing customer service agents with real-time, intelligent voice understanding to powering sophisticated meeting transcription services and improving accessibility features, the model provides a robust foundation. The availability of both streaming and non-streaming options, coupled with competitive pricing, ensures that developers can select the optimal solution for their latency and processing needs, whether for instantaneous voice commands or detailed post-processing of long-form audio.

These developments, alongside other recent advancements like Anthropic's Model Hardware Standard (MHS) for controlling physical machines with AI and Visual Studio Code 1.135's 'Rubber Duck' AI agent for code review, signal a broader industry push towards more agentic, controllable, and integrated AI systems. Developers are increasingly being equipped with AI models that can not only generate content but also understand context, follow complex instructions, and interact with the physical and digital world in more sophisticated ways. This trend is set to accelerate innovation across industries, making AI a more integral and manageable part of the software development lifecycle.