Skip to main content
Capabilities

Model Services

Bailian CLI command reference for full-modal models: text chat, image, video, speech, and vision understanding.

A single command invokes Bailian's full-modal models from the terminal: text chat and Omni, image generation and editing, video generation and editing, speech synthesis and recognition, and vision understanding.
Append --help to any command for full parameter details, e.g. bl image generate --help. Every parameter that accepts a file URL also accepts a local path — the CLI auto-uploads it to temporary storage (valid for 48 hours), so you never need to host files as URLs yourself.

Chat and Omni

CapabilityCommandTypical modelNotesCommon options
Text generationbl text chatqwen3.8-max (default)Flagship model for the agent era; strong at coding, office, and long-horizon autonomy--model model, --system system prompt, --temperature, --max-tokens, --stream streaming, --output json formatted output
Omnibl omniqwen3.5-omni-plusText / image / audio / video understanding and interaction--image / --audio input, --voice output voice, --audio-format, --audio-out save audio, --text-only, --list-voices
bl text chat --message "Explain what a multimodal LLM is in one sentence"

Image generation and editing

CapabilityCommandTypical modelNotesCommon options
Image generationbl image generateqwen-image-2.0Pro text rendering, photorealism, strong semantic adherence--prompt, --model, --size, --async async task
Image editingbl image editqwen-image-2.0Supports multi-image composition--image repeat to merge multiple images, --prompt edit instruction
bl image generate --prompt "A steaming cup of pour-over coffee on a wooden table, morning light" --size 1024x1024

Video generation and editing

CapabilityCommandTypical modelNotesCommon options
Video generationbl video generatewan3.0-video (default)Text-to-video and image-to-video, faithful dynamic motion--prompt, --image first frame (image-to-video), --download <path> save output, --async async task
Reference-to-videobl video refwan3.0-video (default)Up to 9 reference images; stable subject and scene--image reference (repeatable), --image-voice subject voice, --prompt
Video editingbl video edithappyhorse-1.0-video-edit (default)Natural-language video editing--video input, --prompt edit instruction, --download <path> save output
Video downloadbl video downloadDownload a completed video task--task-id (required), --out <path> (required)
Video task querybl video task getQuery async video task status--task-id
bl video generate --prompt "Waves crashing on rocks at sunset, slow camera push-in" --download ./output.mp4

Speech synthesis and recognition

CapabilityCommandTypical modelNotesCommon options
Speech synthesisbl speech synthesizecosyvoice-v3-flashMulti-voice streaming TTS; clone from 5–20s samples--list-voices, --voice, --instruction natural-language style control, --enable-ssml, --stream, --concurrent, --seed reproducible synthesis
Speech recognitionbl speech recognizefun-asr30 languages, including Chinese dialects and accents--diarization speaker separation, --speaker-count, --vocabulary-id hotword list, --language, --async, --out save result, batch up to 100 files
bl speech synthesize --voice longwan --text "Welcome to the Bailian CLI" --stream

Vision understanding

CapabilityCommandTypical modelNotesCommon options
Vision understandingbl vision describeqwen3-vl-plus (default)Long-form video analysis, chart/document parsing, multilingual OCR--image / --video input (local path or URL), multi-image input, --prompt, --output json
bl vision describe --image ./chart.png --prompt "Summarize the key takeaway from this chart"