Paste or drag media into the input box, or drag in previous generation results to use them as references. Right-click a generated image to copy it. Supported media formats and attachment limits depend on the selected model.
Open an independent chat window within a project to select a model, upload attachments, use voice input, search the web, and revisit past conversations. Available to subscribers only. It does not automatically read or modify project content.
Crop images and video frames, trim video or audio, and rotate, flip, remove backgrounds from, or split images into a grid. Both images and videos support color adjustments and LUTs, so you can process existing media directly in the preview.
Describe changes with local annotations, text, and reference media, then reopen the annotations for further editing after generation. Video doodle editing supports videos from 4 to 30 seconds; annotated frames and additional reference images are limited to 30 in total.
Browse and preview content that has not yet been migrated from the legacy Asset Library. Migrate individual items or the entire library and review the results. This entry is available to users with legacy content awaiting migration.
Subscribers can remove subtitles from eligible videos. Generated videos are free to process within 24 hours of generation. Uploaded videos and videos outside the free window require payment based on a quote, with Standard or Professional mode available. The video's longest side must not exceed 2560 pixels.
Legacy source projects and their migrated copies are listed separately and can be deleted independently. Deleting a migrated copy does not delete its legacy source.
Select several media items in the list, navigation panel, or canvas and drag them into an edit together, instead of adding them one at a time.
Create a folder directly from the Move to Folder menu, then select it to complete the move.
Copy personal projects from the new workspace to your team workspace while keeping the originals, so your team can continue working on them.
Choose in settings whether Flova proactively recommends suitable Skills for your current task.
Enable the sandbox in preferences to choose the model it uses. Follow task progress in the conversation. The sandbox is available only to subscribers.
New users can open example projects to explore existing content and learn how creation works in Flova.
Switch between list, grid, and canvas views at the top of your project. The grid supports square cards and masonry layouts, with masonry also available in the Asset Library.
Save creative and collaboration preferences that apply across tasks. Changes take effect with your next request and do not affect tasks already in progress.
Organize assets, documents, and edits with groups and sequences. Switch between list and canvas views, and use the navigation panel to find, select, and move project content.
Create and rename folders, move projects into them, or create projects inside a folder to organize your work.
Open independent conversations within one project and browse message history. Permission modes control which actions need confirmation, and keyboard shortcuts can be customized in settings.
Keep core rules in `skill.md` and supporting material in reference documents. Use multiple Skills in one project. Project documents and Skills support Markdown image previews and editing.
Adjust the Prompt, model, parameters, and reference media manually. Save a draft before generating, or use existing media as a reference for further creation.
Create several edits in one project and switch between them using tabs. Adjust the aspect ratio and select multiple edits for a single export.
Users with eligible content can import legacy projects into the new workspace without overwriting the originals. The Asset Library also supports importing additional legacy content.
Nest groups and sequences as references, detach references, and create copies. Reference library content in project conversations to reuse it across projects.
More Skills load earlier as you scroll, reducing the wait when you reach the bottom of the list.
Saved Skills now show their saved content and reference documents, addressing missing references and inconsistent content when opening details from a direct link.
Usage counts appear beside the author information on cards, in details, and in homepage previews, keeping covers and titles unobstructed.
Links to a specific category open that category in Featured Skills, without requiring you to switch categories manually.
Images generated with Flare or Sunburst can now be doodled on and annotated before generating again.
Flare and Sunburst are available as new image model choices; GPT Image 2 remains available.
Choose Standard or Professional mode for video enhancement, and decide separately whether to use frame interpolation. Professional mode no longer automatically enables frame interpolation.
Drag images from a web page into the input box to upload them, or drop them at a target location in a project. You can also paste images into the project workspace to continue creating.
The team seat limit has increased from 20 to 100.
When generating content, credits expiring within 7 days are used first, followed by credits dedicated to the current model, reducing credits lost to expiration.
If generation pauses because of insufficient credits or restricted model or resolution access, the affected step now shows the specific reason.
When using audio as reference media, M4A, AAC, and FLAC are now supported in addition to WAV and MP3.
Natively generate 30-second videos and create directly from documents and web pages.
Search public Skills within the current category to find suitable creative workflows faster.
Added subtitle removal, video extension, regenerate with original settings, and 1080p output. Extensions and regenerations preserve the original references and settings.
Open separate conversations for shots, characters, audio, and Unsorted Media, then follow up, upload references, and review history while the main Agent continues in parallel.
Reference brand images, documents, videos, and audio while editing a Skill. Published Skills pin those versions and bring usable copies into the user's project.
Codex and other AI tools can connect through MCP authorization to inspect or create projects, upload media, run tasks, restore projects, and export results.
Upload longer scripts, brand materials, research reports, and production notes without splitting them into multiple files first.
Generate up to 30 seconds in one run and add up to 50 multi-modal references across images, videos, and audio for greater control over long narratives and complex scenes.
Images generated with Seedream 5 Pro and GPT Image 2 can be doodled on and annotated directly as reference images to guide regeneration at specific positions and structures.
Let the Agent decide, confirm the Prompt before media generation, or generate directly. Key-step confirmation can be saved per account and project.
Sort by Featured, Most Used, or Newest.
Send more messages while a task is running, edit or delete queued messages, or send one immediately. Failed messages remain available for retry.
Review the exact differences after the Agent edits a document or Skill before saving.
Add details to an existing report and view official replies and unread status without creating another report for the same issue.
Generate multi-modal video with text, image, video, and audio references.
Projects from a personal workspace can now be copied into a team workspace, making it easier to hand off, review, or reuse work that started as an individual project.
Both the project Docs panel and the Asset Library support document upload. Reference material, scripts, brand materials, and other production documents can become long-term context for the Agent inside a project, or be saved to the Asset Library for reuse in later projects.
Timeline audio editing and the Media Space canvas now support more precise operations, including zoom, rotation, flip, and snapping, making media placement, visual adjustment, and audio-video editing easier to control.
Image, video, and audio assets can now be trimmed, with the trimmed result created as a new asset. This makes it easier to extract the exact frame, clip, or sound you need from existing material and continue creating from it.
The chat box now supports voice input, so you can speak creation requests, edit notes, or review feedback directly to the Agent.
The Storyboard canvas now supports Cmd/Ctrl+0 to fit the canvas, +/- to zoom, and Esc to exit or clear selection, with shortcut descriptions and active-scope handling added. Organizing complex storyboards and large sets of media is smoother.
The Skill sharing dialog now includes a six-digit passcode. You can give the passcode to Flova CLI so external Agents such as Codex and Claude Code can reference the corresponding Skill directly, without repeatedly copying the full Skill content.
Video and audio clips now support fade-in duration controls, giving you finer control over how visuals and sound enter a clip.
Project documents and reference materials can now contain longer text, making them better suited for full scripts, brand manuals, client briefs, and production materials.
The underlying Agent creation workflow has been updated to HappyHorse 1.1, continuing to improve project progress and complex task handling.
The account page now supports switching between personal and team workspaces. The team management page supports member lookup and binding, seat renewal, credits allocation and recovery, plus team plan display for managing members and usage in one place.
The previous Storyboard and Media Files Panel are now merged into Media Space. Media Space adds Canvas View, while the original display is renamed Preview List, and the two views can be switched freely. The canvas supports core operations such as asset dragging, preview, zoom, multi-select, and layout saving, making it better for managing story structure and media together in complex projects.
Generation Tools now include the Gemini Omni Flash video model, adding another option for video generation.
Audio generation now supports Seed Audio 1.0, adding a new model option for sound creation.
The lightweight video model Seedance 2.0 Mini has been integrated across frontend and backend. It supports text-to-video, image-to-video, and multi-media-to-video generation, with up to 9 reference images and 4–15 second outputs.
The account page now supports issuing invoices for orders, viewing invoice status, and reissuing invoices. Refunded orders disable the invoice entry, and invoice results are shown with polling updates so the processing state is easier to follow.
You can now let AI Agents such as Codex and Claude Code operate Flova for you. From the Agent access page, copy the starter prompt and complete browser authorization or API Key authorization. The Agent can then create projects, move creation forward, read project status, and export videos, making Flova easier to connect to your own automated creation workflow.
You can save reusable content from a project into the Asset Library and call it again in later projects. Key elements, shot designs, background music, individual media files, and project documents can all become cross-project assets, reducing repeated uploads and cleanup while helping your recurring styles, characters, and production materials build into a reusable creative library.
After a storyboard is generated, you no longer need to return to the main chat and ask the Agent to re-read the whole project for every change. You can comment separately on multiple shots, key elements, or audio layers and let the Agent handle those local adjustments in parallel, so refining shots, character details, and audio-visual pacing can move faster across several edit points at once.
You can extract the audio track directly from a video asset preview and create a new audio asset. The extracted audio can be reused for voiceover, audio reference, or editing, making sound material from existing videos easier to reuse.
Skills can now be discovered and reused from the community. You can browse selected Skills, review the workflows that fit your needs, and add them to My Skills for direct use in projects. Common creation methods, style rules, and production workflows no longer need to be rebuilt each time, and it is easier to start from a mature template.
When a creative task needs more background, factual detail, or reference material, the Agent can run a search as part of its plan and organize the results into a document resource inside the project. Later creation steps can keep using those results, giving scripts, settings, and visual plans a stronger information base.
Before generating an image or video, the Agent can create an editable prompt draft and keep it as a draft card in the media area. You can open the draft to check the Prompt, reference media, and generation settings, then save and generate; drafts also support deletion, moving, undo/redo, and version-conflict handling, so you can refine the Prompt before spending credits.
Project documents can now be created, edited, renamed, and deleted as freeform documents, and the Agent can also write intermediate work into project documents during creation. For example, you can ask the Agent to draft a script, character notes, or shooting instructions first, then keep editing them in the Docs panel; uploaded documents can also be previewed in the Docs panel and referenced directly as Agent context.
The promo code and invite history modules have been reopened. The reward dialog now shows the current reward tiers and a complete invite history including invitees' emails; the list auto-loads the next page as you scroll near the bottom. If your account doesn't yet qualify to invite — for example less than 7 days old, or no project has been successfully exported yet — the copy-link button is disabled and a tooltip explains why. Credits earned through referrals now clearly show "valid for 30 days" in the hint, so the expiration is easy to see.
Voiceover now supports MiniMax, adding over a thousand new voices and a richer set of emotion parameters. Pick them straight from the voice selector in the voiceover dialog; the default narration model has also switched to MiniMax.
Voiceover generation now has a Speed parameter so you can dial pacing up or down. It covers all voiceover providers (Doubao, ElevenLabs, MiniMax), and the chosen speed is saved with the asset, so you don't have to re-tune it next time you reuse the same voiceover.
From the asset preview area you can now upscale images and videos with AI. Images can go up to 4K or 8K with a choice of magnification levels; videos can go up to 1080p, 2K, or 4K with 24, 30, or 60 fps options. After you click AI Upscale, the parameter panel shows the cost for the current job, which you can confirm before running. The feature is available to paid subscribers; users without a subscription see an unavailable state on the entry. Failed jobs can be retried in one click.
Generation Tools also has an AI Upscale entry now, so a freshly generated image or video can be upscaled in place without going back to the asset preview. The resulting upscaled version is easy to locate from history, and a "view original" entry lets you compare the original and upscaled side by side.
A Feedback entry has been added to the avatar dropdown in the top-right corner. Clicking it opens a feedback form where you can write a description and contact info to send to us directly. The form has validation and duplicate-submit protection, with a confirmation prompt once sent.
Splitting audio and video clips on the timeline has been polished overall: a dedicated split icon, the clip stays selected after a split, and hover hints help you cut precisely; split audio keeps a continuous fade-out, and the fade can be adjusted independently; split operations support undo and redo, and edge cases where a split overlaps with a trim are handled more reliably.
Assets in the Media Files Panel and the storyboard's asset area now support multi-select — you can box-select or check several assets at once to favorite, delete, or run other right-click operations in batch. You can also drag files directly from your desktop into either area to upload, without opening the file picker first.
When generating images or video, you can now have the AI draft the generation prompt first, preview and edit it, and only trigger the actual generation when you're satisfied. This lets you confirm the prompt matches your intent before spending credits, so iterating on the prompt no longer consumes compute each time. When you duplicate a project, drafted prompts come along to the new copy too.
The homepage video wall is now organized into content channels by category, so you can browse inspiration by genre. Every video card is also linked to its corresponding Skill: open the fullscreen preview to view the Skill's details and add it to your My Skills (or remove it again) in one click. Videos without a custom cover automatically use their first frame as the preview image, keeping the wall visually consistent.
When the Agent asks you to clarify a request during creation, the question is now shown as a set of clickable option cards instead of one long inline question. Both single- and multi-select are supported, and each option includes a short description, so it's easier to understand what each choice means before you pick.
While a storyboard is generating, the "restore to a previous version" button is automatically disabled, preventing a rollback before the new result arrives and avoiding version conflicts.
Skills you create can now be shared publicly through a link. Anyone who opens the link can preview the Skill's full rules right away without signing in; the login prompt appears only when they decide to save it to their own library. Skills publicly shared by others in the community can be added to your My Skills with a single click, ready to use in your next project. To stop sharing, a confirmation dialog appears before the share is revoked, so you don't undo it by accident.
The Kling video model is upgraded to 3.0 with a new All-Purpose Reference mode — when generating a video you can supply multiple reference images and a reference video at the same time, used together as combined guidance for style, framing, and motion. When you switch models, the allowed image count, video duration, and upload rules are automatically re-aligned and filtered for the current model, so parameters from the previous model don't leak through.
Flova's biggest release since launch — the in-project workspace has been fully rebuilt, Skill is introduced for the first time, and existing projects can be migrated to the new format in one click.
The in-project workspace has been rebuilt around four independent panels — Storyboard, Media Files, Timeline, and Docs — that you can switch between and operate one at a time.
Many operations that used to be Agent-only are now available to you directly: drag-and-drop for assets, attaching media to the storyboard and individual shots, editing documents and text content, deleting items one by one, and more.
Flova introduces Skill for the first time — a reusable creation workflow for the Agent. You can capture your habits, taste, and production rules as a Skill so the Agent works to the same standard every time.
Projects created before 1.0 can be synced to the new format in one click, so you can keep working in the new four-panel workspace without redoing them.
