# Changelog All notable changes to PixVerse CLI will be documented in this file. ## [1.4.15](https://github.com/PixVerseAI/cli/releases/tag/v1.4.15) — 2026-10-27 ### Features - Add `kling-4.0` video model (`Kling 4.0`) — available in `create video`, `create reference`, and two-frame `create transition`, with 720p / 1080p output (default 1080p), 3–30s durations, automatic image-to-video framing, and up to 10 image / 5 video references (3–30s per video, 30s total) ## [1.4.14](https://github.com/PixVerseAI/cli/releases/tag/v1.4.14) — 2026-10-27 ### Bug Fixes - Fix media upload stability ## [1.4.13](https://github.com/PixVerseAI/cli/releases/tag/v1.4.13) — 2026-10-23 ### Bug Fixes - Allow `canvas patch apply` and `canvas patch dry-run` to submit patches when `base_edit_version` differs from the latest Graph version. Preserve the supplied version and all other input and contextual validation. ## [1.4.12](https://github.com/PixVerseAI/cli/releases/tag/v1.4.12) — 2026-10-23 ### Bug Fixes - Fix media upload stability ## [1.4.11](https://github.com/PixVerseAI/cli/releases/tag/v1.4.11) — 2026-10-23 ### Features - Add `--relax-mode [true|false]` / `--no-relax-mode` to `create image`, `create video`, `create transition`, `create reference`, `create modify`, and `create motion-control` — enable with the bare flag or `true`, disable with `false` or `--no-relax-mode`, pass choices through to the backend, and omit `relax_mode` when neither flag is supplied, with model support and account entitlement validated by the backend - Add `--background` to GPT Image 2.5 Sunburst and Flare in `create image`, with `auto` (default), `transparent`, and `opaque` backgrounds for both text-to-image and image-to-image generation ## [1.4.10](https://github.com/PixVerseAI/cli/releases/tag/v1.4.10) — 2026-10-21 ### Features - Add MCP-to-CLI pairing ## [1.4.9](https://github.com/PixVerseAI/cli/releases/tag/v1.4.9) — 2026-10-15 ### Features - Extend `create template` to support more template types, with ordered image, video, and text inputs, template-defined output settings, and polling for returned image/video results ## [1.4.8](https://github.com/PixVerseAI/cli/releases/tag/v1.4.8) — 2026-10-09 ### Features - Add `gemini-nano-banana-2.1` image model (`Nano Banana 2.1`) — available in `create image` in the global region, with 1080p / 1440p / 2160p output, Auto and fourteen fixed aspect ratios, selectable Detail Level via `--detail-level` (`minimal` / `medium` / `high`, default `medium`), and up to 14 reference images ## [1.4.7](https://github.com/PixVerseAI/cli/releases/tag/v1.4.7) — 2026-10-08 ### Bug Fixes - Fix media upload stability ## [1.4.6](https://github.com/PixVerseAI/cli/releases/tag/v1.4.6) — 2026-09-22 ### Bug Fixes - Remove local `create music` prompt and lyrics length limits, deferring provider-specific validation to the backend - Remove local `create voice` text length limits, allowing provider-specific validation errors to pass through from the backend ## [1.4.5](https://github.com/PixVerseAI/cli/releases/tag/v1.4.5) — 2026-09-20 ### Features - Add `canvas arrange ` to automatically arrange and save an entire Canvas project's node layout, with positive decimal ID validation, pure JSON output, and guidance for reading the updated layout - Accept positional project IDs across all 15 project-scoped Canvas operations, including graph reads, node versions, patches, and dispatch; retain `--project-id` compatibility, preserve large IDs as strings, and reject conflicting inputs before making requests ## [1.4.4](https://github.com/PixVerseAI/cli/releases/tag/v1.4.4) — 2026-09-15 ### Features - Add `minimax-h3-max` video model (`MiniMax H3 Max`) — available in `create video`, `create reference`, and two-frame `create transition`, with 480p / 768p / 1080p output, 5–15s durations, automatic image/reference framing, and up to 9 image / 3 video / 3 audio references (12 total) ## [1.4.1](https://github.com/PixVerseAI/cli/releases/tag/v1.4.1) — 2026-09-09 ### Features - Add `gpt-image-2.5-sunburst` (`GPT Image 2.5 Sunburst`) and `gpt-image-2.5-flare` (`GPT Image 2.5 Flare`) to `create image`, with 1080p / 1440p / 2160p generation, ten supported aspect ratios at every quality, `low` / `medium` / `high` / `xhigh` / `max` detail levels, and up to 16 reference images; `gpt-image-2.5-flare` is now the default image model ## [1.4.0](https://github.com/PixVerseAI/cli/releases/tag/v1.4.0) — 2026-09-07 ### Features - Discover the installed CLI's Create capabilities offline with `capabilities` and `capabilities create`, including supported modes, model-specific parameters, defaults, enums, ranges, units, and media limits, without signing in or making a network request - Query current Canvas generation routes, models, supported inputs, and parameter constraints with `capabilities canvas`; filter by node type, route, or model, or return the original Canvas definitions with `--raw` - Add the top-level `canvas` command group for building and managing connected Canvas generation workflows from the CLI - Create empty Canvas projects directly with `canvas project create`, with optional name and description fields and active workspace support - Inspect Canvas projects with `canvas graph`, including nodes, dependencies, edit versions, generation states, and validation issues, and inspect individual node schemas with `canvas node schema` - Plan and apply Canvas changes safely with `canvas patch dry-run` and `canvas patch apply` — read a JSON literal, local file, or stdin, validate changes before saving, retry them safely, and reject unsupported dependency edits before they are saved - Control generation precisely with `canvas dispatch`, confirm exact node batches with `canvas dispatch rebind`, and recover selected nodes with `canvas graph reconcile` - Manage the full Canvas node lifecycle — inspect nodes, browse and apply saved versions, rerun generation, and extract a video's complete audio track into a Canvas audio node - Use Canvas with the CLI's existing login, pure JSON output through `--json` / `-p`, and standard exit codes for reliable agent automation ## [1.3.10](https://github.com/PixVerseAI/cli/releases/tag/v1.3.10) — 2026-09-04 ### Documentation - Update public README ## [1.3.6](https://github.com/PixVerseAI/cli/releases/tag/v1.3.6) — 2026-09-03 ### Features - Add `flux-3.0` (`FLUX 3`) for text-to-video and image-to-video creation, with 720p/1080p output, 5–20s durations, selectable Auto and fixed aspect ratios, optional generated audio, and the required FLUX safety parameter - Add `wan-3.0` (`Wan 3.0`) for video, two-frame transition, and mixed-media reference creation, with 480p/720p/1080p output, 2–30s durations, optional generated audio, audio-only references, and up to 10 images / 5 videos / 5 audios (20 references total) ## [1.3.5](https://github.com/PixVerseAI/cli/releases/tag/v1.3.5) — 2026-08-19 ### Features - Add `1080p` quality support to Seedance 2.5 across `create video`, `create reference`, and two-frame `create transition`, while keeping `720p` as the default and all existing duration, aspect-ratio, and reference limits unchanged ## [1.3.4](https://github.com/PixVerseAI/cli/releases/tag/v1.3.4) — 2026-08-18 ### Features - Add `music-3.0` (`MiniMax Music 3.0`) and `music-v2` (`ElevenLabs Music V2`) to `create music` — both support lyrics, auto lyrics, instrumental generation, and 10–240s durations, while `music-2.6` remains the default ## [1.3.3](https://github.com/PixVerseAI/cli/releases/tag/v1.3.3) — 2026-08-17 ### Features - Add `--task-type ` to Seedance 2.5 `create reference` — accepts `auto` (default), `reference`, `edit`, or `extend` and sends the selected task intent as `task_type`, with local model and value validation before media resolution - Add high-resolution external video tiers — `kling-o3-4k` supports video, two-frame transition, and mixed-media reference creation as a separate 4K model tier; `kling-3.0-4k` supports video and two-frame transition as a separate 4K model tier; Kling requests omit the quality parameter; and Grok Imagine 1.5 image-to-video accepts 1080p ### Bug Fixes - Make Grok Imagine `create reference --videos` submit automatic duration (`duration: 0`), omit the derived source-video aspect ratio, and reject fixed video durations, while keeping image-only references on the normal fixed-duration path - Make Gemini Omni `create reference --videos` default and lock duration to `auto`, submitting automatic duration and rejecting fixed values while keeping image-only references on the normal fixed-duration path ## [1.3.2](https://github.com/PixVerseAI/cli/releases/tag/v1.3.2) — 2026-08-09 ### Features - Expand `create reference` with model-specific video editing — V6 accepts up to 10 images / 2 videos, Gemini Omni accepts up to 5 images / 1 video, Kling O3 accepts up to 7 images or 4 images with 1 video, and Grok Imagine accepts either images or 1 video, with local validation for each model's media constraints - Add `--duration auto` handling for video references — V6 locks video inputs to automatic duration, while Seedance 2.5 defaults video inputs to automatic duration but allows an explicit 4–30s value - Add `--aspect-ratio auto` to Seedance 2.5 text-to-video and reference creation — automatic reference duration locks the aspect ratio to `auto`, while fixed duration keeps automatic and fixed ratios selectable ## [1.3.1](https://github.com/PixVerseAI/cli/releases/tag/v1.3.1) — 2026-08-07 ### Features - Add `seedance-2.5` video model (`Seedance 2.5`) — available in `create video`, `create reference`, and two-frame `create transition`, with 480p / 720p, 4–30s durations, selectable fixed aspect ratios for video/reference, and up to 30 image / 10 video / 10 audio references (50 total; video and audio each ≤ 30s) ## [1.3.0](https://github.com/PixVerseAI/cli/releases/tag/v1.3.0) — 2026-08-04 ### Features - Add a top-level `miniapps` command group for PixVerse MiniApps (preset generators): `miniapps list` shows the available apps as an ID/Name/Description table, `miniapps info ` shows an app's details and its normalized parameter schema (required fields, types, enum values, and a copy-pasteable example), and `miniapps create --id --params ` submits the app's args as required JSON (a literal string, a file path, or `-` for stdin) and returns a `project_id` - Extend `task` and `asset` with `--type miniapps` to query, wait on, download, and delete MiniApp projects by their `project_id` ## [1.2.13](https://github.com/PixVerseAI/cli/releases/tag/v1.2.13) — 2026-08-04 ### Bug Fixes - `create template` now supplies a random seed for image templates when `--seed` is omitted, preventing submissions from failing with `invalid param` ## [1.2.12](https://github.com/PixVerseAI/cli/releases/tag/v1.2.12) — 2026-08-03 ### Features - Add `768p` quality to the `minimax-h3` video model alongside the existing `1440p` option ### Bug Fixes - `create reference` now preserves explicit fixed `--aspect-ratio` values for `minimax-h3` image references, while continuing to default to `auto` when omitted ## [1.2.11](https://github.com/PixVerseAI/cli/releases/tag/v1.2.11) — 2026-07-31 ### Features - Add `minimax-h3` video model (`MiniMax H3`) — available in `create video`, `create reference`, and two-frame `create transition`, with 1440p, 5–15s durations, and up to 9 image / 3 video / 3 audio references ## [1.2.10](https://github.com/PixVerseAI/cli/releases/tag/v1.2.10) — 2026-07-14 ### Features - Improve `create upscale` with a `2160p` default and broader video input support — `--video` now accepts local files, HTTPS URLs, video IDs, and media paths - Allow `task status` to query multiple space-separated IDs in parallel while retaining the existing `--ids` syntax - Add exit code `7` (`CONCURRENCY_LIMIT`) for temporary concurrency limits, allowing agents to retry them separately from credit errors ### Bug Fixes - Make batch status exit codes deterministic when requests fail with different error types - Fix the workspace help text to reference the `status` subcommand ## [1.2.9](https://github.com/PixVerseAI/cli/releases/tag/v1.2.9) — 2026-07-12 ### Bug Fixes - `create music` no longer rejects `--auto-lyrics` on `lyria-3-pro-preview` (Google Lyria 3 Pro) — every music model supports auto lyrics - `create music` flag combinations now match the web app: `--instrumental` takes precedence (forces `auto_lyrics` off and omits lyrics), and explicit `--lyrics` is dropped when `--auto-lyrics` is on ## [1.2.8](https://github.com/PixVerseAI/cli/releases/tag/v1.2.8) — 2026-07-10 ### Features - Add `seedream-5.0-pro` image model (`Seedream 5.0 Pro`) — available in `create image`, with 1080p / 1440p generation, standard aspect ratios, and up to 10 reference images ## [1.2.7](https://github.com/PixVerseAI/cli/releases/tag/v1.2.7) — 2026-07-03 ### Features - Add `gemini-omni-flash` video model (`Google Gemini Omni`) — available in `create video` and `create reference`, with 720p, 3–10s durations, `16:9` / `9:16`, and a 5-image reference cap - Add `gemini-3.1-flash-lite` image model (`Nano Banana 2 Lite`) — supports 1080p generation and up to 14 reference images ## [1.2.6](https://github.com/PixVerseAI/cli/releases/tag/v1.2.6) — 2026-06-28 ### Bug Fixes - Fix the CLI failing to start when launched via the installed `pixverse` command or `npx` (affected 1.2.0–1.2.5) ## [1.2.5](https://github.com/PixVerseAI/cli/releases/tag/v1.2.5) — 2026-06-24 ### Features - Add `seedance-2.0-mini` video model (`Seedance 2.0 Mini`) — available everywhere `seedance-2.0-fast` is supported (`create video`, `create reference`, `create transition`) ## [1.2.4](https://github.com/PixVerseAI/cli/releases/tag/v1.2.4) — 2026-06-24 ### Features - Add `--audios` to `create reference` for `seedance-2.0` audio references (max 3 clips, each 2–15s, total ≤ 15s, ≤ 15MB per clip; requires at least one image/video reference) ## [1.2.3](https://github.com/PixVerseAI/cli/releases/tag/v1.2.3) — 2026-06-23 ### Features - Add `2160p` (4K) quality to the `seedance-2.0-standard` model (`--quality 2160p`) — available everywhere the model is supported (`create video`, `create reference`, `create transition`) ## [1.2.2](https://github.com/PixVerseAI/cli/releases/tag/v1.2.2) — 2026-06-18 ### Bug Fixes - `create reference` no longer rejects `seedance-2.0` video references when their first frame can't be extracted — frame extraction is now best-effort (previously such inputs failed with "Failed to extract first frame") - Correct the ElevenLabs Music model ID to `music-v1` (previously the invalid `music_v1`) ## [1.2.1](https://github.com/PixVerseAI/cli/releases/tag/v1.2.1) — 2026-06-15 ### Bug Fixes - `create reference` now allows up to 9 reference images on `seedance-2.0` models (previously capped at 7 for all models) ## [1.2.0](https://github.com/PixVerseAI/cli/releases/tag/v1.2.0) — 2026-06-12 ### Breaking Changes - Remove `create speech` (Lip Sync) ### Features - Add `create voice` for text-to-speech, with a read-only `voice` group (`voice models`, `voice presets`) listing available models, languages, and preset voices - Add `create music` for prompt-to-music generation, with a read-only `music` group (`music models`) listing available models - Voice and music outputs are saved as audio assets — list, inspect, and download them with `asset` and `task` - Add `grok-imagine-1.5` video model for image-to-video (`create video --model grok-imagine-1.5 --image `) - Ship a compact, machine-readable `capabilities.json` manifest — commands, flags, exit codes, and the JSON output contract — for agent and Skill consumption ## [1.1.12](https://github.com/PixVerseAI/cli/releases/tag/v1.1.12) — 2026-06-04 ### Bug Fixes - Fix media path validation rules ## [1.1.11](https://github.com/PixVerseAI/cli/releases/tag/v1.1.11) — 2026-06-01 ### Features - `--prompt` and `--tts-text` accept `-` to read the value from stdin (pipe long or multi-line prompts) - Add `pixverse update` to upgrade the CLI, with a daily update-available notice - Add global `--trace-id` option to attach a caller-supplied UUIDv4 to all API requests ## [1.1.10](https://github.com/PixVerseAI/cli/releases/tag/v1.1.10) — 2026-05-25 ### Features - `create modify` support `--images` for reference images ## [1.1.9](https://github.com/PixVerseAI/cli/releases/tag/v1.1.9) — 2026-05-15 ### Features - Add `v6` model support to `create reference` - `create image` default model is now `gpt-image-2.0` ## [1.1.8](https://github.com/PixVerseAI/cli/releases/tag/v1.1.8) — 2026-05-14 ### Breaking Changes - Remove `create sound` command - Drop deprecated models from selected creation modes (passing them as `--model` now fails validation): - `create video`: `v5.5`, `v5`, `v5-fast` - `create extend`: `v5.5`, `v5` - `create reference`: `v5` - `create transition` (2 frames): `v5.5`, `v5`, `v4.5` - `create transition` (3+ frames): `v4.5` ### Features - Add `2160p` quality for `seedream-5.0-lite` - `create --no-wait --json` output now includes the resolved `params` ## [1.1.7](https://github.com/PixVerseAI/cli/releases/tag/v1.1.7) — 2026-05-09 ### Features - Support Seedance 2.0 mixed image+video reference - Support more GPT Image 2 aspect ratios ### Refactor - Simplify GPT Image 2 aspect ratio validation ## [1.1.6](https://github.com/PixVerseAI/cli/releases/tag/v1.1.6) — 2026-04-27 ### Features - Enable grok-imagine for extend and reference modes ## [1.1.5](https://github.com/PixVerseAI/cli/releases/tag/v1.1.5) — 2026-04-27 ### Features - Add happyhorse-1.0 video model ## [1.1.4](https://github.com/PixVerseAI/cli/releases/tag/v1.1.4) — 2026-04-22 ### Features - Add gpt-image-2.0 image model ## [1.1.3](https://github.com/PixVerseAI/cli/releases/tag/v1.1.3) — 2026-04-20 ### Bug Fixes - Auto-compress oversized local images and harden JSON/download behavior ## [1.1.2](https://github.com/PixVerseAI/cli/releases/tag/v1.1.2) — 2026-04-16 ### Features - Add 1080p support for seedance-2.0-standard video generation ## [1.1.1](https://github.com/PixVerseAI/cli/releases/tag/v1.1.1) — 2026-04-16 ### Features - Add veo-3.1-lite model ## [1.1.0](https://github.com/PixVerseAI/cli/releases/tag/v1.1.0) — 2026-04-12 ### Features - Add seedance-2.0 and seedance-2.0-fast models - Add kling-o3 and kling-3.0 models - Add asset list --source/--off-peak filtering and asset upload command with unified upload flow - Add saved folders command with list/items/new/rename/add/remove/delete ## [1.0.13](https://github.com/PixVerseAI/cli/releases/tag/v1.0.13) — 2026-04-10 ### Features - --image and --video support more input types for better compatibility ## [1.0.12](https://github.com/PixVerseAI/cli/releases/tag/v1.0.12) — 2026-04-08 ### Features - Add pixverse-c1 model support, use friendly model display names ## [1.0.11](https://github.com/PixVerseAI/cli/releases/tag/v1.0.11) — 2026-04-07 ### Bug Fixes - Support batch polling for --count > 1 across all create commands ## [1.0.10](https://github.com/PixVerseAI/cli/releases/tag/v1.0.10) — 2026-04-04 ### Features - Add create motion-control command ## [1.0.9](https://github.com/PixVerseAI/cli/releases/tag/v1.0.9) — 2026-04-02 ### Features - Add create modify command with keyframe support ## [1.0.8](https://github.com/PixVerseAI/cli/releases/tag/v1.0.8) — 2026-04-02 ### Bug Fixes - Add `--no-audio` flag to `create video`, `create transition`, `create extend`, and `create reference` to allow disabling audio generation - Add `--no-multi-shot` flag to `create video` to allow disabling multi-shot mode ## [1.0.7](https://github.com/PixVerseAI/cli/releases/tag/v1.0.7) — 2026-03-31 ### Features - Add workspace support (list, status, switch, manage) ## [1.0.6](https://github.com/PixVerseAI/cli/releases/tag/v1.0.6) — 2026-03-30 ### Bug Fixes - Validate image model before API call to prevent silent fallback ## [1.0.5](https://github.com/PixVerseAI/cli/releases/tag/v1.0.5) — 2026-03-30 ### Features - Add V6 model support, refactor defaults caching ## [1.0.4](https://github.com/PixVerseAI/cli/releases/tag/v1.0.4) — 2026-03-23 ### Documentation - Rewrite public README with full product intro, usage guide, and accurate command examples ### Features - Update creation success message to "Generation Task Created" ### Miscellaneous - Update dependencies and configuration files ## [1.0.0](https://github.com/PixVerseAI/cli/releases/tag/v1.0.0) — 2026-03-06 Initial release of PixVerse CLI. ### Creation - **Text to Video / Image to Video** — generate videos from text prompts or reference images (`create video`) - **Text to Image / Image to Image** — generate images from text or transform existing images (`create image`) - **Transition** — create smooth transitions between keyframes (`create transition`) - **Speech** — add lip-sync speech to videos via TTS or audio file (`create speech`) - **Sound Effects** — add AI-generated sound effects to videos (`create sound`) - **Extend** — extend video duration (`create extend`) - **Upscale** — upscale video resolution (`create upscale`) - **Reference** — generate videos with character reference images (`create reference`) ### Task & Asset Management - Check task status and wait for completion (`task status`, `task wait`) - List, inspect, download, and delete generated assets (`asset list`, `asset info`, `asset download`, `asset delete`) ### Account - OAuth device flow authentication (`auth login`) - View account info and credit usage (`account info`, `account usage`) - Open subscription page in browser (`subscribe`) ### CLI Features - Interactive wizard mode for all creation commands - JSON output (`--json` / `-p`) for scripts and AI agents - Configurable defaults (`config set`, `config list`, `config reset`) - Structured exit codes for error handling