Gemini Omni
Follow Google’s Gemini Omni Flash from its first rollout in Flow and the Gemini app to API access, frame control, and longer clips. Compare Google, Runway, and Higgsfield on the same dates.
- Posts archived
- Attachments examined
- Capabilities with evidence
What changed, when
Announcement dates in KST · Select a change or open its example.
Workflow threads
Compare workflows
Compare product availability and limits
Generate & frame
- Google · · Set first and last frames in Gemini API or Flow.
- Runway · · Animate from a first and optional last frame at 3–15 seconds.
- Higgsfield · · Frames mode takes a start and optional end frame.
- Recraft Studio · · The Recraft Studio picker exposes Omni Flash 1.2 to generate video with native audio.
Use references
- Google · · Google improves adherence to reference material in 1.2.
- Runway · · Text generation can use five image and three video references.
- Higgsfield · · Use up to six image and three video references for appearance.
- Recraft Studio · No matching evidence in this archive.
Edit & extend
- Google · · Extend an existing scene and control its continuation.
- Runway · · Choose reference, edit, or extend video-to-video modes.
- Higgsfield · · Extend an existing video up to 30 seconds total.
- Recraft Studio · No matching evidence in this archive.
Draft & deliver
Workflows over time
Announcement dates in KST. Earlier notices establish what was already available.
Update archive
Official titles with edited summaries and post-level sources.
Official examples and illustrations
Play video in its card or enlarge the example for its source and context.
Read the dated updates and sources
- · Omni Flash brings conversational video to Google Flow
Generate video from mixed inputs and revise shots with a prompt. The launch also previews the model in creative tools such as Flow Music. Google introduced a model family; video is the demonstrated output here. Some Flow surfaces were rolling out.
- · Google shows successive edits of the same clip
Google shows reference-driven outputs and a multi-step violin edit that changes the instrument and camera framing over several prompts. The post publishes curated examples and describes rollout to eligible Gemini and Flow subscribers.
- · Omni Flash reaches the Gemini API in public preview
The API adds generation and conversational editing from text, images, and video, with output up to ten seconds. At this point the model did not correctly process video references, and uploading an audio reference or extending a scene was unsupported.
- · Runway Dev adds Omni Flash
Generate 3–10 seconds from text or a first image, or edit an existing clip at 720p with up to five image references. Edits preserve the input clip’s duration and orientation. Generated audio is optional.
- · Omni 1.1 adds keyframes, scene extension, and 4K
Use first and last frames, extend a clip, combine video references, and iterate at 360p before upscaling selected shots. The release separates Gemini API controls from Google Flow controls; delivery and input settings differ by mode.
- · Flow adds frame controls and draft-to-final export
Set start and end frames, make low-cost 360p drafts, and export at higher quality including 1080p or 4K. Flow lists these as product features for the launch day.
- · Higgsfield adds frame-to-frame transitions and draft mode
Set both frames, generate the transition, and test at 360p before rendering at up to 4K. The product notice lists a ten-second maximum per generation.
- · Runway Dev adds Omni 1.2 with edit and extend modes
Generate 3–15 seconds from text, use a first and optional last frame, or use video-to-video reference, extend, and edit modes. Text-to-video accepts up to five image and three video references; output goes from 360p to 4K with mode and input charges.
- · Omni 1.2 expands generation to 15 seconds
Google improves adherence to references and multi-turn editing, adds up to 15 seconds of native generation, and reduces per-second pricing. Google lists availability in the Gemini API, Google Vids, and Gemini for business. Mode-specific input and output rules apply.
- · Higgsfield adds Omni 1.2 References, Frames, and Extend
Choose native-audio generation with references or frames, or continue an existing clip. Reference mode accepts up to six images and three videos. Reference videos provide appearance; their audio is ignored. Extend can add 3–10 seconds per step up to 30 seconds total.
- · Recraft adds Gemini Omni Flash 1.2 with native audio
Create up to ten seconds of video with native audio. Recraft highlights stronger subject continuity and clearer multi-speaker lip sync. Recraft Studio requires a plan with video and external models. Its ten-second limit is for this product surface.
Documentation and availability notes
- Runway Dev API · · Runway’s model directory lists the original and 1.2 variants and their respective model IDs and supported interfaces.
- Google Gemini API · · Operational constraints depend on the model and the mode. The model family announcement and every third-party integration have their own release dates.